The GEO Manifest 2025: Complete Guide to Generative Engine Optimization

1. From SEO to GEO – The Evolution of Search

Introduction: For over two decades, “search” has been practically synonymous with Google. The act of Googling became the default way people found information online ( ^[1] ). In that time, Search Engine Optimization (SEO) rose to dominate digital marketing, as businesses large and small vied for visibility in those all-important search results. But today, the search landscape is shifting under our feet. A new breed of AI-powered answer engines – led by generative models like ChatGPT – is changing how people seek answers. Users are increasingly turning from traditional search engines to AI chatbots that deliver instant, conversational responses. This chapter examines how we got here, why generative AI is disrupting search as we know it, and what Generative Engine Optimization (GEO) means for marketers aiming to stay visible in this new era. We’ll explore the legacy of SEO, the seismic impact of AI on search behavior, define GEO in a marketing context, and explain why optimizing for AI-generated answers has become mission-critical in 2025.

1.1 The Legacy of SEO

SEO has been the backbone of online marketing for the past twenty years. Since the early 2000s, ranking high on search engine results pages (SERPs) – especially on Google – meant a flood of organic traffic and potential customers. In fact, an estimated 68% of online experiences begin with a search engine , underscoring how crucial search visibility has been for businesses ( ^[2] ). From finding a new restaurant to researching B2B software, users overwhelmingly turned to Google’s results, and companies invested heavily to appear at the top. Over the past 15+ years, SEO became a cornerstone of digital marketing , credited with powering startups to massive growth. One analysis notes that SEO-centric strategies helped numerous startups and scaleups reach billion-dollar valuations ( ^[3] ). The formula was straightforward in concept: understand what keywords your audience is searching, create relevant content, and optimize your website’s technical factors and backlinks so that Google’s algorithm ranks you highly for those searches. Achieving a top organic ranking on a high-volume query could transform a business’s fortunes. Unlike paid ads, these clicks were “free” (after the upfront content and optimization effort), making SEO a high-ROI strategy for those who mastered it. The dominance of SEO in driving traffic is evident from industry data. Even as of a few years ago, organic search was the single largest driver of website traffic across most industries. For example, BrightEdge research found that in 2019, organic search accounted for 53% of trackable website traffic on average, far more than social media or other channels ( ^[4] ). Marketers often quoted “Page 1 or bust,” emphasizing that if you aren’t on the first page of Google results, your content might as well be invisible. This pressure to rank led to an entire ecosystem of SEO professionals, agencies, and tools devoted to understanding Google’s ever-changing ranking algorithms. The SEO industry ballooned into a multi-billion dollar sector . By recent estimates, companies worldwide were spending nearly $80 billion on SEO by 2023 , up from $47 billion in 2020 ( ^[5] ). This spend includes everything from hiring SEO experts and creating optimized content to investing in platforms like Moz, SEMrush, or Ahrefs for analytics. The rationale was clear: as long as search engines drove the lion’s share of web traffic, investing in SEO meant tapping into a huge audience. Over time, SEO best practices evolved (from keyword-stuffing in the early days to focusing on high-quality content and user experience today), but the end goal remained the same – get to the top of Google’s results . It’s hard to overstate Google’s dominance in this era. Google became the default homepage of the internet for many. Capturing one of the top organic slots on a popular Google query could yield thousands or millions of monthly visitors. Marketers optimized for Google above all else, since Google has long held around 90%+ of the search engine market share globally ( ^[6] ). Microsoft’s Bing and others lagged far behind. Naturally, search engine optimization in practice largely meant Google optimization. This legacy of SEO is one of continuous adaptation. Google constantly tweaked its algorithms (with major updates like Panda, Penguin, Hummingbird, RankBrain, and countless core updates), keeping SEO practitioners on their toes. Yet, despite these shifts, the fundamental premise held: if you publish valuable, relevant content and earn authority (through quality backlinks and references), your site can rank well and attract visitors from organic search. SEO became a standard discipline in marketing departments , alongside advertising and PR. For many businesses, organic search traffic became a key performance indicator (KPI) and a reliable source of leads or revenue. However, even before AI chatbots arrived, cracks in the traditional SEO model had begun to appear. Google’s own changes started to reduce organic visibility – for instance, more ads and rich features now occupy the top of many search results, pushing organic links further down ( ^[7] ). By the late 2010s, Google was answering more queries directly on the results page (through Featured Snippets, Knowledge Panels, etc.), leading to a rise of “zero-click searches.” Studies found that over 60% of Google searches now end without any click to a website – users get their answer from Google’s snippet itself ( ^[8] ). Even a #1 ranked page doesn’t guarantee a flood of traffic like it once did; the top result’s average click-through rate has fallen to around 27%, with the second result around 8% ( ^[9] ). This trend of users not needing to click a result was an early sign that the search landscape was shifting. Still, those snippets often pulled from websites (for example, a featured snippet quoting a line from a blog), so SEO experts adapted by optimizing for Answer Boxes and snippets – a tactic dubbed Answer Engine Optimization (AEO). In essence, SEO was already evolving to feed answers directly to users via Google’s interface. In summary, the legacy of SEO is one of immense impact on digital marketing. It dominated the last two decades as the primary way to drive online visibility ( ^[10] ). Companies that cracked the SEO code reaped enormous rewards in traffic and sales, fueling the growth of an $80B industry. But just as marketers thought they had mastered the art of pleasing Google’s algorithms, a new disruption has emerged – one that doesn’t play by the old rules of web search.

1.2 AI Disrupts Search

A major upheaval began in late 2022 with the advent of powerful generative AI chatbots. OpenAI’s ChatGPT burst onto the scene and within months reached viral adoption. In fact, ChatGPT’s user growth was unprecedented – Morgan Stanley research noted it hit 100 million monthly active users faster than any consumer application in history ( ^[11] ). Practically overnight, millions of people started asking ChatGPT all sorts of questions, from coding help and academic research to advice and general knowledge. For the first time, a significant number of users were getting direct answers from an AI instead of typing queries into Google. This shift has been swift. By early 2023, the buzz around AI chatbots grew so loud that Google’s management reportedly issued a “code red.” Internal teams were told to treat the rise of ChatGPT as an existential threat to Google’s core search business ( ^[12] ). (To put that business in perspective: Google’s search advertising revenue was about $149 billion annually ( ^[13] ), so anything that might pull users away from Google search was a huge concern.) Google wasn’t alone in noticing the trend – Microsoft invested heavily in OpenAI and quickly integrated GPT-4 into Bing , launching a new AI-enhanced Bing Chat in early 2023. For the first time in years, Google Search had a compelling alternative nipping at its heels, offering conversational answers integrated with search results. What’s driving this AI-fueled disruption? User experience. AI chatbots can deliver one-stop answers in a conversational format, which many users find more convenient than clicking through multiple links. For example, when someone asks ChatGPT “best project management tools for startups,” they receive a concise recommendation list with explanations – not a cluttered page of 10 blue links ( ^[14] ). The AI filters out the noise and presents what it “thinks” is the best answer, often drawing from information across numerous sources. This means the user spends less time hunting and more time getting insights. It’s a fundamentally different paradigm of information retrieval – more of a dialogue than a traditional query-and-results process. Real-world data confirms that people are indeed shifting some of their search behavior to AI platforms. One major survey in early 2025 found that 27% of Americans now use AI chatbots like ChatGPT instead of search engines for at least some queries ( ^[15] ). In the UK, similar numbers were reported, reflecting a broad international trend. And among younger, tech-savvy users, the preference for AI answers can be even higher. Another study by Elon University in 2025 revealed that half of U.S. adults have used an AI system like ChatGPT, Google’s Gemini, Anthropic’s Claude, or similar – and two-thirds of those users treat these tools like search engines to find information ( ^[16] ) ( ^[17] ). In other words, a sizable chunk of the population has already begun searching via chat interfaces. Users cite the speed and convenience of getting direct, context-rich answers. Rather than comb through several webpages, they can ask an AI and get a synthesized answer that feels tailored to their question ( ^[18] ). Generative AI tools like ChatGPT are rapidly becoming alternative gateways to information, often bypassing traditional search engines entirely. A 2025 survey found that more than a quarter of Americans now turn to AI chatbots (such as ChatGPT, Bing Chat, or Google’s Bard) instead of using Google for certain queries ( ^[15] ). This shift indicates that search behavior is moving from keyword queries toward conversational interactions, cracking the dominance of the classic “Google search” experience. Crucially, it’s not just ChatGPT in the game. Following OpenAI’s breakthrough, Google launched its own generative AI search initiatives . In mid-2023, Google unveiled the Search Generative Experience (SGE) – an AI-powered summary that appears at the top of search results for certain queries, synthesizing information with cited sources. They also rolled out Google Bard , a conversational AI, and invested in a next-gen language model called Gemini to power future search features. By late 2024, Google’s AI Overviews (the result of these efforts) were reaching over 1 billion users and expanding globally ( ^[19] ). Google is effectively augmenting its familiar search results with AI-generated answers, so that users might not need to click external sites for many questions. Microsoft’s Bing, meanwhile, integrated OpenAI’s GPT-4 into Bing’s interface and saw a surge of interest, albeit from a smaller base. Other tech players jumped in as well – Meta released open-source LLMs like LLaMA that others could adapt, Anthropic pushed its Claude chatbot known for a very large context window, and in late 2023 xAI (Elon Musk’s AI venture) debuted “Grok” , an AI assistant with a bit of a personality. Even Apple signaled moves in this space, announcing that AI-driven search engines (like a ChatGPT integration) would be offered as options via Siri ( ^[20] ). In short, the search market that was once basically Google, with a dash of Bing/Yahoo, is now fragmenting into a plethora of AI-driven alternatives . Some of these new AI search entrants are particularly noteworthy. Perplexity AI , for instance, launched as an AI-centric search engine that delivers answers with cited sources. It gained attention by providing a conversational interface with transparency about where information is coming from (something ChatGPT lacked initially). By early 2025, Perplexity’s usage was climbing rapidly – their daily active users grew 800% year-over-year to 15 million ( ^[11] ) – indicating real demand for AI-enhanced search tools. Even OpenAI introduced a version of “ChatGPT search” (sometimes referred to as SearchGPT ) that allowed the chatbot to browse the web and present sources. In one month in late 2024, OpenAI’s own search referral traffic (people clicking out from ChatGPT’s answers to external websites) jumped 44%, while Perplexity’s grew 71% ( ^[21] ). These are small players compared to Google, but they’re growing fast. BrightEdge data noted that by 2025, ChatGPT’s built-in search functionality was on pace to capture about 1% of total search market share (in terms of query volume), which might sound tiny but would equate to a $1.2+ billion shift in value if realized ( ^[22] ). All of this has profound implications for the $80B SEO ecosystem. As Andreessen Horowitz (a16z) quipped in a 2025 analysis, “The foundation of the $80 billion+ SEO market just cracked.” ( ^[20] ) The traditional model – optimize for Google, get traffic, get conversions – is being upended. If users no longer need to click links (or if they’re not even using Google to begin with), then the old playbook needs rewriting. We are essentially witnessing the end of “search as we knew it.” The familiar “ten blue links” on a Google results page are giving way to what one expert calls “direct answer engines” ( ^[23] ). In these new AI-powered search experiences, the engine doesn’t just rank websites, it composes an answer drawing from its trained knowledge or real-time web data. The user’s attention goes to the AI’s response, not necessarily to a specific website. If the AI’s answer happens to cite sources or suggest follow-up links, those become bonus visibility opportunities – but they are not guaranteed. From a consumer standpoint, search is becoming more of a conversation. Instead of formulating the perfect keyword query, users can ask questions in natural language and even ask follow-ups for clarification. The AI will remember the context (to an extent) and refine answers. This conversational search approach was accelerated by ChatGPT’s mainstream moment ( ^[14] ). Soon after, we saw multi-turn dialogue capabilities integrated into Bing and Google’s search UIs as well. The result is that the barriers between “searching” and “chatting” are blurring – a revolution in how we discover information. For online marketers, this is a disruptive change. The traffic streams long taken for granted may dwindle if more people get what they need from an AI’s answer. Early evidence of impact can be seen in consumer behavior shifts: ask yourself, how often do you now receive an answer directly from Siri, Alexa, or an AI chatbot, without needing to scroll a website? That frequency is only increasing. Marketers have noticed instances of declining click-through rates when Google displays an AI-generated summary on top – users might get their answer from the summary and not scroll further. If this trend continues, optimizing solely for classic web search will no longer be enough .

1.3 Defining Generative Engine Optimization (GEO)

The rise of AI-driven answers has given birth to a new concept: Generative Engine Optimization (GEO) . Just as SEO is all about making your content visible and highly ranked on search engines, GEO is about making your content visible and favored by generative AI answer engines . In other words, how do you ensure that when an AI like ChatGPT, Bard (Gemini), Claude, or Bing Chat is answering questions in your domain, it includes your brand’s knowledge, content, or offerings in its response? GEO is the discipline of optimizing content for AI systems that generate responses, rather than simply for indexing and ranking links on a page ( ^[24] ). Let’s unpack that. A traditional search engine (Google) uses a crawler and an index, matches keywords to documents, and ranks those documents (webpages) based on relevance and authority signals. If you did good SEO, your webpage might rank #1 and users clicking that result would see your content. A generative AI, on the other hand, might have read your webpage during its training or via real-time search, and when asked a question, it might pull information from it – not by showing the user your webpage, but by weaving the info into its own answered text . The user might not ever visit your site or even know the answer came from you unless the AI explicitly cites sources (and not all do, or they might cite only a few of many sources). Optimizing for these AI platforms requires new strategies . It’s about ensuring the AI knows about your content and trusts it enough to use in answers , and ideally that it cites or attributes your brand so the user is aware of your contribution. In practical terms, GEO involves several facets, many of which are still being figured out by the industry. At its core, it means creating content and site structures that AI can easily ingest, understand, and regurgitate accurately . Some aspects include: using clear, structured formats (so that facts or instructions from your content can be extracted correctly), maintaining strong credibility (so that AI models consider your content authoritative and worth including), and providing fresh, up-to-date information (since AI systems, especially those with real-time retrieval, favor current data for many queries ( ^[25] )). GEO also means considering how an AI prioritizes information . Unlike Google’s algorithm, which heavily weights backlinks and keywords, an AI model might prioritize contextual relevance or factual accuracy . For instance, AI engines often cite official studies, data, or highly authoritative sources even if those didn’t rank high in Google ( ^[26] ). This suggests that in the GEO era, creating genuinely authoritative content (research, whitepapers, expert insights) can pay off even more, as the AI may latch onto those facts. Another difference is personalization across AI systems. Each AI platform has its own way of sourcing and presenting info. As one analyst observed, “Every AI model prioritises different signals. ChatGPT values semantic depth, Perplexity emphasizes structured data, [Google] Gemini takes a trust-first approach. Optimising for AI feels like targeting a million different search engines.” ( ^[27] ) In SEO we mainly optimized for one dominant algorithm (Google’s). In GEO, we might need to consider multiple AIs – e.g., ensuring our content is formatted with schema markup and sources (helpful for Bing/Perplexity which value citations), written with rich context and detail (helpful for ChatGPT’s style), and backed by trustworthy reputations (helpful for Google’s SGE/Bard). This fragmentation adds complexity: GEO isn’t a single algorithm to game, but a spectrum of AI behaviors to accommodate. So how do we practically define success in GEO? The basic idea is your brand becomes part of the AI-generated answer . Suppose someone asks an AI, “What are the top security software for small businesses?” Under the old model, success is if your company’s website ranks on page 1 of Google for “security software small business.” Under GEO, success might look like: the AI’s answer text says, “You should consider XYZ Security Suite (by YourCompany) as it’s highly rated for small businesses…” or the AI cites YourCompany’s blog as a source of expert information on the topic. In either case, the user is exposed to your brand within the answer itself . It’s a win even if they don’t immediately click a link, because your brand entered their consideration via the AI’s recommendation. This is fundamentally what GEO aims to achieve – to maintain brand visibility when the “answers” are coming from an algorithmic interlocutor instead of a list of web links . It’s worth noting that GEO builds on some concepts from earlier in the search evolution. A few years back, marketers started talking about Answer Engine Optimization (AEO) – optimizing for featured snippets and voice assistant answers (like Siri/Alexa responses). GEO is like AEO on steroids, expanded to all generative models and chat-style interactions. With voice search and snippet optimization, you were still often trying to get a specific bite-sized answer from your site to be showcased (for example, a definition or step-by-step list). GEO encompasses that but also the broader goal of being woven into AI’s knowledge base . It’s not always a one-shot factual snippet; it could be being referenced as an authority or having your data used in an AI’s explanation. We are still in the early days of formalizing GEO tactics, but it’s clearly emerging as a parallel to SEO. In May 2025, one industry commentator declared, “We’re witnessing the birth of Generative Engine Optimization – a new discipline that will be as important as SEO was for the last two decades.” ( ^[10] ) Forward-thinking marketing teams are already experimenting with techniques: injecting likely Q&A prompts into their content, publishing FAQ-style pages that match conversational queries, ensuring their content is included in public knowledge bases (like Wikipedia or scholarly databases) that AI models are known to scrape, and even using the AI tools themselves to audit whether and how their brand appears in answers ( ^[28] ). We’ll delve into specific strategies in later chapters, but the key point here is that GEO is about influence without a click. It asks: Will the AI mention or recommend us? – a very different question from Can we rank and get a click? The concept of Generative Engine Optimization (GEO) centers on ensuring your content is favored by AI-driven answer engines. Rather than competing for a top spot on a search results page, GEO is about earning a place in the AI’s generated response itself ( ^[29] ). This means creating authoritative, well-structured content that conversational AI systems trust and cite. As AI search grows, GEO is becoming the new frontier of digital visibility, where the goal is to have your brand’s information seamlessly woven into the answers that chatbots and AI assistants provide.

1.4 Why GEO Matters Now

The urgency around GEO is growing by the day. We are witnessing a rapid pivot in how consumers find information – a pivot that could leave traditional SEO-reliant strategies in the dust if businesses don’t adapt. Here are several key reasons why GEO demands attention right now : Users are bypassing traditional search : As discussed, a significant and growing percentage of users are going straight to AI assistants for answers. If one-quarter or more of searches shift to AI by 2025–2026 (as Gartner predicts about 30% by 2026 ( ^[30] )), that’s a huge chunk of queries where the Google SERP might not even be seen. If your marketing strategy today is 100% based on getting clicks from Google, you risk losing reach as those eyeballs move to chat-based interfaces. GEO matters because it’s the way to regain visibility in those interactions. It’s no longer hypothetical – over 50% of U.S. adults have now used an AI chatbot, and many use them like search tools ( ^[16] ) ( ^[17] ). This trend will only accelerate as AI becomes integrated into smartphones, virtual assistants, and everyday apps. The rise of zero-click answers : Even within traditional search engines, the trend is toward answers without clicks . Google’s introduction of features like featured snippets, Knowledge Graph panels, and now AI Overviews means users can often get what they need without leaving Google at all. By 2024, about 60% of Google searches ended without any click to external content ( ^[8] ). With AI-generated answers (which are even richer and more comprehensive than a simple snippet), the proportion of zero-click experiences could climb higher. For marketers, this means that even if people use Google, they might not be visiting your site unless your information is part of Google’s answer. GEO is about making sure your information is that answer . It’s a direct response to the zero-click phenomenon – if the user won’t click out, you need your brand to live inside the answer unit . New traffic and branding pathways : While an AI answer might not send a click, it can send customers . For example, if an AI assistant recommends a particular product (say, “The user asks: what is the best noise-cancelling headphone?” and the AI responds with a short list of models), being named in that list is invaluable brand exposure. A user might then directly navigate to your site or search your brand name later – an attribution path that is hard to track but very real ( ^[31] ). This is what some have dubbed the “attribution black hole” of AI: customers influenced by AI recommendations who convert later via direct or brand search traffic. GEO matters because it positions your brand to be in those crucial AI-driven recommendations. In essence, **GEO can drive implicit traffic and demand, even if it’s not the classic referral click . Ignoring it means ceding those invisible opportunities to competitors. Competitors (including new ones) are seizing AI visibility : One startling observation in the post-ChatGPT world is that the brands dominating AI answers aren’t always the ones that dominated SEO. Unknown or niche competitors who have content structured well for AI can suddenly surface in answers where bigger SEO players are absent ( ^[32] ). If your competitor’s content is being used by the AI to formulate answers and yours isn’t, they effectively become the voice that consumers hear. Early adopters of GEO tactics are likely already gaining an edge. This is especially true in industries where factual data or community Q&A content (like StackExchange, Wikipedia, etc.) feed the AI – a smaller player who happens to be active on those platforms might be all over AI answers. GEO is the proactive response to ensure you don’t become invisible in the new landscape while agile competitors leapfrog you in AI presence . AI integration is everywhere : Beyond just search engines, consider how generative AI is creeping into many consumer touchpoints. AI chat is being built into smart home devices, office productivity tools, customer service bots, and more. If a user asks their AI-powered personal assistant, “Book me a flight and hotel for Athens next weekend,” and the assistant uses generative AI to recommend options, travel brands will want to have their offerings included. This extends to virtually every domain – personal AI assistants could soon handle a large share of queries that traditionally would be web searches** ( ^[30] ). GEO principles (making sure the AI has your data, and cites your services) will apply in all these contexts. In short, the future of discovery isn’t just a search engine results page; it’s ambient AI-driven discovery . Ensuring your content is AI-friendly now is what will keep you discoverable as this future arrives. Trust and accuracy considerations : It’s worth noting that AI answers, while convenient, bring along issues of accuracy (hallucinations) and bias. As a brand, one might fear being misrepresented by an AI or omitted due to some quirk of the model’s training data. Engaging in GEO isn’t just an opportunity; it’s also a defensive move. By actively optimizing and providing reliable, well-structured information, you increase the chances that AIs will pull correct info about your brand (rather than outdated or incorrect data). For instance, if your product pricing or specs aren’t clearly available in structured form, an AI might scrape a third-party site and give a wrong answer. GEO involves feeding the AI ecosystem with accurate data about your brand – through schema markup, data feeds, or updated web content – to mitigate misinformation. Given that generative models sometimes “make up” answers, having your authoritative content in their training or retrieval sources is crucial to ensure factual representation. In summary, GEO matters now because user behavior has already changed and the platforms are quickly following. The search and discovery ecosystem in 2025 is not the one we knew in 2020. We’re at an inflection point where AI-driven answers are becoming mainstream, and businesses must adapt or risk losing visibility. As one SEO expert put it, the task ahead is not just to capture traffic from search engines, but to become the source that AI engines cite and recommend ( ^[33] ). That is the essence of GEO. It’s the next chapter in the evolution of search marketing, picking up where traditional SEO leaves off. Companies that recognize this shift early and integrate GEO strategies stand to maintain and even grow their reach in the age of AI. Those that don’t may find themselves witnessing their hard-earned Google rankings bringing diminishing returns, as the audience simply bypasses the old pathways. Closing thought: The evolution from SEO to GEO represents a fundamental change in how we approach online visibility. SEO isn’t dead – far from it. Traditional search still drives massive value, and Google isn’t going away overnight ( ^[6] ). But the playing field is expanding . It now includes AI chatbots, voice assistants, and other generative experiences as additional “front doors” through which customers will find information. Marketers need to broaden their optimization efforts accordingly. The following chapters will dive deeper into how SEO is converging with new practices like Answer Engine Optimization and Large Language Model Optimization, the inner workings of LLMs like ChatGPT and Google’s Gemini, and practical strategies to thrive in this new environment. The key takeaway for now is: search is no longer just about being ranked, it’s about being referenced . GEO is about ensuring that when the answers are generated, your brand’s voice is part of the conversation ( ^[14] ) ( ^[29] ). The evolution of search has begun – and it’s time for marketers to evolve with it. Sources: The insights and data in this chapter were drawn from a range of up-to-date sources, including industry surveys, expert analyses, and reports as of 2024–2025. Key references include BrightEdge’s research on AI-driven search growth ( ^[34] ) ( ^[35] ), a Tom’s Guide survey on user preferences shifting to AI chatbots ( ^[15] ), an Elon University study on LLM adoption ( ^[16] ), as well as commentary by thought leaders in SEO and AI (Andreessen Horowitz, Rand Fishkin, Kalyani K.) highlighting the disruption and opportunities at the intersection of search and generative AI ( ^[20] ) ( ^[10] ). These will be further elaborated in subsequent chapters as we explore tactical approaches to GEO in depth.

References

^[1] www.tomsguide.com - Tom's Guide URL: https://www.tomsguide.com/ai/new-study-reveals-people-are-ditching-google-for-the-likes-of-chatgpt-search-heres-why

^[2] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/seo-statistics

^[3] CustomerThink Article - CustomerThink URL: https://customerthink.com/seo-in-2025-navigating-a-changing-landscape

^[4] www.brightedge.com - BrightEdge URL: https://www.brightedge.com/blog/organic-share-of-traffic-increases-to-53

^[5] CustomerThink Article - CustomerThink URL: https://customerthink.com/seo-in-2025-navigating-a-changing-landscape

^[6] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

^[7] CustomerThink Article - CustomerThink URL: https://customerthink.com/seo-in-2025-navigating-a-changing-landscape

^[8] CustomerThink Article - CustomerThink URL: https://customerthink.com/seo-in-2025-navigating-a-changing-landscape

^[9] CustomerThink Article - CustomerThink URL: https://customerthink.com/seo-in-2025-navigating-a-changing-landscape

^[10] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[11] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[12] Undark.Org Article - Undark.Org URL: https://undark.org/2023/01/19/google-search-has-nothing-to-fear-from-chatgpt

^[13] Undark.Org Article - Undark.Org URL: https://undark.org/2023/01/19/google-search-has-nothing-to-fear-from-chatgpt

^[14] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[15] www.tomsguide.com - Tom's Guide URL: https://www.tomsguide.com/ai/new-study-reveals-people-are-ditching-google-for-the-likes-of-chatgpt-search-heres-why

^[16] www.elon.edu - Elon University URL: https://www.elon.edu/u/news/2025/03/12/survey-52-of-u-s-adults-now-use-ai-large-language-models-like-chatgpt

^[17] www.elon.edu - Elon University URL: https://www.elon.edu/u/news/2025/03/12/survey-52-of-u-s-adults-now-use-ai-large-language-models-like-chatgpt

^[18] www.tomsguide.com - Tom's Guide URL: https://www.tomsguide.com/ai/new-study-reveals-people-are-ditching-google-for-the-likes-of-chatgpt-search-heres-why

^[19] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

^[20] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[21] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

^[22] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

^[23] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[24] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[25] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[26] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[27] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[28] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[29] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[30] www.tomsguide.com - Tom's Guide URL: https://www.tomsguide.com/ai/new-study-reveals-people-are-ditching-google-for-the-likes-of-chatgpt-search-heres-why

^[31] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[32] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/80b-search-industry-being-rewritten-most-startups-dont-kalyani-khona-cqr6f

^[33] www.genmark.ai - Genmark AI URL: https://www.genmark.ai/resources/blog/generative-engine-optimization

^[34] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

^[35] www.globenewswire.com - Globe Newswire URL: https://www.globenewswire.com/news-release/2024/12/16/2997560/0/en/New-Report-from-BrightEdge-Reveals-Surge-in-AI-Search-Engines-Signaling-a-New-Era-in-Online-Information-Discovery.html

2. Key Concepts – SEO vs. AEO vs. GEO vs. LLMO

Chapter Overview: In this chapter, we break down four key acronyms that represent the evolution of search optimization in the age of AI. We’ll clarify the definitions and goals of each concept – SEO, AEO, GEO, and LLMO – and discuss how they overlap and differ. These concepts all address the same fundamental challenge (gaining visibility when users seek answers), but each has a distinct focus and set of strategies. A solid grasp of these will help online marketing professionals adapt to AI-driven search and maintain their brand’s visibility.

2.1 SEO – Search Engine Optimization

Definition: Search Engine Optimization (SEO) is the traditional practice of optimizing a website to rank higher on search engine results pages (SERPs) like Google or Bing. The goal is to increase organic (non-paid) visibility for relevant queries, thereby driving more clicks and traffic to your site. SEO focuses on aligning your content and website attributes with the search engines’ ranking algorithms, emphasizing keywords, backlinks, and technical best practices to improve relevance and authority. In simpler terms, SEO is about making your content easily discoverable by search engines for the terms people use, and convincing the engines that your content is the most authoritative answer for those terms. This involves several interrelated components: On-Page Optimization: Researching keywords that your audience searches for and incorporating them naturally into your content (titles, headings, body text). Ensuring the content thoroughly addresses the intent behind those keywords. Structuring pages with clear headings and using meta tags (title, description) that signal relevance to the query. In 2024, Google’s algorithms are highly sophisticated at interpreting natural language, so SEO has evolved from just “keyword stuffing” to focusing on topics , entities , and satisfying user intent in content. Off-Page and Backlinks: Building the authority of your site via backlinks – links from other reputable websites. High-quality backlinks act as endorsements, telling search engines that your content is trustworthy and valuable. A strong backlink profile has long been a cornerstone of SEO success. For instance, content that attracts links from news outlets or industry sites tends to rank higher, as Google sees those links as votes of confidence. Technical SEO: Ensuring the website is technically sound so search engine crawlers can easily find and index your content. This includes having a logical site architecture, clean URLs, fast loading speed , mobile-friendly design, and proper HTML markup. Technical SEO also covers things like fixing broken links, creating sitemaps, and using schema markup (structured data) to help search engines understand your content context. A technically healthy site is more likely to be crawled and ranked appropriately. Context & Importance: SEO has been the backbone of digital marketing for decades. As of 2024, Google still dominates search with about ~90% of global market share ( ^[1] ), handling over 13 billion searches per day ( ^[2] ). (For perspective, one analysis found ChatGPT at around 1 billion interactions per day – growing fast but still far behind Google’s volume ( ^[2] ).) This sheer volume of searches means high rankings on Google can funnel tremendous traffic to a business. Even as new AI tools emerge, traditional search remains a primary way consumers find information and products. For example, Google processes over 99,000 searches every second (roughly 8.5 billion searches per day ) ( ^[3] ). Each search is an opportunity for a brand to connect with a user. SEO is about capturing those opportunities by appearing on page one, ideally at the top of the results for relevant queries. Core SEO Strategies and Metrics: Classic SEO success is measured by metrics like your keyword rankings , organic traffic from search, click-through rate (CTR) on your snippets, and ultimately conversions (e.g. leads or sales from organic visitors). If you rank #1 for a valuable query, you’ll get a large share of clicks – historically, the top result gets ~20-30% of clicks or more for a given search. Thus, SEO efforts often prioritize achieving those top positions. Many of the best practices in SEO also improve user experience: fast, mobile-friendly pages and high-quality content benefit both search rankings and user satisfaction. Google’s quality guidelines emphasize E‑E‑A‑T – experience, expertise, authoritativeness, and trustworthiness – for content creators ( ^[4] ). Ensuring your site demonstrates these qualities (for instance, by having expert authors, citing reputable sources, and providing accurate, useful information) can boost both SEO and user trust. Evolution: It’s worth noting that SEO is not a static field. Search algorithms update frequently (Google alone rolls out thousands of small updates and several major updates each year). In recent years, AI has already been influencing traditional search : Google uses AI models like BERT and MUM to better understand natural language queries and content, and Bing has integrated OpenAI’s GPT-4 into its search interface for more conversational answers ( ^[5] ). However, these advancements have been happening under the hood of search engines. The fundamental output of SEO – i.e., a ranked list of links on a SERP – remained the same, until very recently. SEO thus provides the foundation upon which newer concepts like AEO, GEO, and LLMO build. If SEO’s mantra was “be visible in search results,” the following evolutions extend that mantra to new formats and new platforms for visibility.

2.2 AEO – Answer Engine Optimization

Definition: Answer Engine Optimization (AEO) is an evolution of SEO that focuses on optimizing your content to be directly delivered as answers by search platforms, rather than just appearing as one link among many. In other words, instead of merely aiming for a high rank, AEO strives to make your content the actual answer snippet that a search engine or digital assistant provides to the user. This concept emerged as search engines (and voice assistants) began providing quick, concise answers (often called “featured snippets” or answer boxes ) at the top of search results, and as users increasingly use voice queries expecting spoken answers. AEO emphasizes capturing “position zero” – the featured snippet spot above the traditional results – and dominating in voice search responses. It recognizes that modern users often prefer immediate answers without clicking through, especially for simple queries. For example, if a user asks, “What is the capital of Italy?” Google might show a box at the top with the answer “Rome” (often extracted from a site like Wikipedia), and voice assistants like Alexa or Google Assistant will speak that answer aloud. AEO is about structuring your content so that your site provides that answer . Key aspects of AEO include: Featured Snippets Optimization: Featured snippets are the boxed answers on Google’s results (also known as “instant answers”). They can be paragraphs, lists, tables, or videos, extracted from a web page that Google deems to best answer the question. Winning a featured snippet can dramatically increase visibility. In fact, featured snippets account for an estimated 35% of all clicks on Google searches ( ^[6] ) – a testament to how many users are drawn to that instant answer. AEO tactics involve formatting content to directly answer likely questions in a concise way (roughly 40–60 words for paragraph snippets) and using clear structures (Q&A formats, lists, step-by-steps) that Google can easily pull from ( ^[7] ) ( ^[8] ). Figure 2.1: Example of a Google featured snippet (answer box) showing a direct answer at the top of the search results ( ^[6] ) ( ^[9] ). Featured snippets grab attention by providing concise, authoritative answers instantly. Being the source of such an answer not only drives some traffic, but also builds brand trust by positioning you as the authority on the question. Voice Search Optimization: With the proliferation of smart speakers and voice assistants (Siri, Google Assistant, Alexa, etc.), many queries are now spoken aloud. Voice searches tend to be longer and in natural language , often very specific questions (e.g. “OK Google, what’s the best Italian restaurant near me that’s open now?”). AEO involves capturing those voice query answers. Typically, voice assistants draw answers from featured snippets or knowledge graph data. If your content is optimized to be a featured snippet, it’s more likely to be read aloud as the voice answer. The importance of voice search is growing – for example, voice commerce was projected to reach about $80 billion in annual value around 2025 ( ^[10] ), illustrating how significant voice interactions have become for businesses. To optimize for this, content creators use conversational keywords (phrases that sound like how people speak) and implement schema like Speakable (a markup that indicates sections of text suited for text-to-speech) ( ^[8] ). Direct Answer Content Strategy: AEO encourages structuring content to think in questions . Instead of just presenting information, you frame headings as likely user questions and immediately answer them. For example, an FAQ section on a product page is very AEO-friendly: the question is a heading (“How long does the battery last?”) and the answer is a succinct paragraph right below. Similarly, a blog post might include a section titled “What are the benefits of XYZ?” followed by a bullet-point list or brief answer. This Q&A formatting helps search engines easily identify potential answers for commonly asked questions ( ^[7] ). Tools like Google’s People Also Ask box, Answer the Public, or BuzzSumo’s Question Analyzer can help discover what questions people are asking in your domain ( ^[11] ) – which you can then target in your content. Structured Data & Knowledge Graph: AEO often leverages structured data (using Schema.org markup) to make answers more accessible to search engines. For instance, implementing FAQ schema on your FAQ page can explicitly tell Google the Q&A pairs on that page. Using HowTo schema can highlight step-by-step instructions. These not only boost chances of appearing as rich results or snippets, but also feed into Google’s Knowledge Graph for factual queries ( ^[12] ) ( ^[13] ). The Knowledge Graph is Google’s internal knowledge base that supplies quick facts (like definitions, dates, etc.). While you can’t directly control the Knowledge Graph except through being a trusted source or providing data, using schema and being cited by other trusted sources can increase the chance that your information is what powers those instant answers. Why AEO Matters Now: Over the past few years, user search behavior has shifted significantly toward quick-answer retrieval. Studies show a growing prevalence of “zero-click searches,” where the user’s query is answered on the results page itself, so they have no need to click any result. One analysis in 2020 found nearly 65% of Google searches ended without a click because Google provided an answer or other immediate result on the page ( ^[14] ). This trend has likely continued into 2024. For businesses, this means if you’re not the featured answer, you might get zero visibility or traffic from certain queries, since the user got what they needed without ever visiting a website. AEO is essentially a response to this phenomenon – ensuring that if an answer is going to be shown directly, it comes from your content. Additionally, the explosion of AI-driven search and chat (discussed further under GEO) makes AEO even more critical. As an example, by early 2023 when OpenAI’s ChatGPT became mainstream, some websites saw a decline in their search traffic because users found answers via AI. Stack Overflow (a Q&A site for programmers) reported an 18% drop in visits after ChatGPT’s rise, as developers got code answers directly from AI instead of visiting the site ( ^[15] ). This is essentially an “answer engine” (ChatGPT) diverting traffic. On the flip side, companies that embraced answer-focused content saw benefits: NerdWallet (a finance advice site) experienced a 35% growth in revenue despite a 20% decline in site traffic, by ensuring their content and brand expertise still reached consumers through snippets and other answer platforms ( ^[16] ). In other words, even if fewer people clicked through, NerdWallet’s authoritative answers kept influencing users (perhaps via featured snippets, voice answers, etc.), leading them to trust and use NerdWallet when it came time to make decisions. This underscores a key AEO insight: Success isn’t just about clicks; it’s about presence in the answer ecosystem. AEO vs. Traditional SEO: It’s important to note that AEO doesn’t replace SEO; it builds upon it . In fact, a strong SEO foundation (technical health, good content, authority) is often a prerequisite for effective AEO ( ^[17] ). Many tactics overlap – for example, page speed and crawlability (SEO basics) also help AEO because if your content isn’t indexed or loads slowly, it won’t be used in snippets. However, there are some strategic shifts: Goal Difference: Traditional SEO’s goal is to get the click – to entice the user to click your link on the SERP. AEO’s goal is to satisfy the query right on the SERP or via voice – even if no click occurs ( ^[18] ) ( ^[19] ). With AEO, you accept that the user might get their answer and not visit your site, and that’s okay as long as your brand delivered the answer. This means success metrics for AEO include things like number of featured snippets captured, voice query share, and brand mentions, not just site traffic ( ^[20] ). You might track, for instance, how often your domain is the source of a Google snippet, or how often a voice assistant cites your brand (“According to Example.com…”). Content Style: SEO content might be lengthy and comprehensive (to cover a topic in depth), and while that’s still valuable, AEO demands that within that content you extract the concise answer . AEO best practice is to put the direct answer in the first 1-2 sentences of a section, then elaborate with details ( ^[21] ). For example, an SEO-driven blog on “how to improve credit score” might be 2,000 words. To optimize for AEO, you might begin a section with “ How can you improve your credit score? You can improve it by consistently paying bills on time, reducing your debt, avoiding new credit inquiries, and correcting any errors on your credit report.” – a 1-2 sentence summary answer – and then follow with more explanation or steps. This way, Google can grab that summary for a snippet, and the user can read on if they want more. Technical & Structured Data: As noted, AEO places extra emphasis on schema markup like FAQPage, HowTo, etc., and on semantic HTML structure (proper use of headings, lists, tables) so that answer engines can easily “parse” your content ( ^[22] ) ( ^[7] ). If SEO is about making your page rank, AEO is about making the content itself easily extractable. In summary, AEO is about being the answer . It recognizes that search is no longer just a referral service sending visitors to websites; increasingly, the search engine itself wants to deliver the information. To remain visible, organizations must adapt by feeding the answer engines the content they need, in the format they prefer. Many forward-thinking businesses have added AEO to their content strategy – monitoring featured snippet opportunities, crafting Q&A-driven content, and even creating content specifically to address common voice queries. This lays the groundwork for the next step: optimization for generative AI answers.

2.3 GEO – Generative Engine Optimization

Definition: Generative Engine Optimization (GEO) is a emerging discipline that extends the principles of SEO/AEO to a new class of platforms: AI-driven answer engines powered by large language models . In simpler terms, GEO is the process of ensuring your content is visible and favored within AI-generated responses – for example, the answers produced by ChatGPT, Bing’s AI chat, Google’s Search Generative Experience (SGE) “AI overviews,” or conversational search engines like Perplexity. The acronym “GEO” encapsulates the idea that we now have generative engines (AI systems that generate answers) alongside traditional search engines, and we need to optimize content for those generative engines ( ^[23] ). GEO positions your brand to appear when users ask questions to AI platforms. This might mean having your content explicitly cited as a source (as Bing Chat and Perplexity often do), or at least having the substance of your content woven into the AI’s answer. The ultimate goal of GEO is similar to SEO: increase your visibility and attract targeted traffic – except the traffic might come through an AI intermediary. It’s about engaging potential customers where they are , even if that’s in a chatbot conversation rather than a search results page ( ^[24] ). Examples of AI “Generative Engines”: To ground this, consider some prominent generative AI search tools in 2024–2025: OpenAI ChatGPT: The AI chatbot that ignited the mainstream AI frenzy. Users can ask ChatGPT anything and get a composed answer. By default, ChatGPT’s free version has a knowledge cutoff (currently 2021), but the paid and enhanced versions (or with plugins) can browse current info or use retrieval. ChatGPT typically does not cite sources in its answers (unless a plugin or browsing is used), but it’s trained on vast web data. If your content was in its training data, it might influence ChatGPT’s answers. GEO aims to ensure that when ChatGPT is asked about your domain, it reflects information from your content – ideally even mentions your brand if appropriate. Microsoft Bing + GPT-4 (Bing Chat): Microsoft integrated OpenAI’s GPT-4 model into Bing Search. When users search on Bing, they have the option to engage a chat mode that answers in a conversational style and does cite sources . For instance, Bing Chat might answer a complex query with a few paragraphs and footnote each sentence with links to the website it came from. As of 2024, Bing had only ~3–4% search market share globally ( ^[25] ), but it saw a surge of interest when it launched the AI features (Bing app downloads quadrupled after introducing AI chat ( ^[26] )). GEO for Bing means you’d want your content to be one of those cited links in a Bing Chat answer. In practice, strong traditional SEO helps here – Bing’s AI often pulls from the top search results. So if you rank well on Bing, you’re more likely to be referenced by the AI. Google’s Generative Search (SGE “AI Overviews”): Google has been testing and rolling out an experimental feature in Search called the Search Generative Experience (SGE) , which produces an “AI overview” at the top of the results for certain queries. This overview is a few bullet points or sentences synthesized from various sources, with links to those sources (often 2-3 pages) shown beside the answer. Essentially, Google is doing what Bing did – using an AI to summarize information for the user. Google calls these “AI overviews” or “AI snapshots” . By late 2024, Google began rolling out AI Overviews to broad audiences in the US ( ^[27] ). Optimizing for this means ensuring your content can be picked up by Google’s AI as part of the summary. According to Google and SEO experts, strategies include using clear factual statements the AI might quote, maintaining good SEO (so your page is among the top results considered), and providing schema/structured data which Google can trust. In essence, GEO for SGE overlaps a lot with AEO: if you provide concise answers and have high authority, you increase the chance Google’s AI overview will draw from your site. (Google has also been working on their LLM called Gemini – launched in late 2023 – which is expected to power Bard and search features, making them more multimodal and powerful to compete with GPT-4 ( ^[28] ) ( ^[29] Perplexity AI: Perplexity is an AI search engine specifically designed to answer questions with large language models + live web data , and it always provides citations . If you ask Perplexity a question, it will generate an answer like ChatGPT but with footnote numbers that link to sources (often showing text excerpts). For marketers, Perplexity represents an ideal scenario: the AI not only uses your content but also directly links to it, potentially driving traffic. GEO strategy for Perplexity would involve appearing in the top search results for the query (since it often pulls from the first page of Google/Bing) and having content that a language model finds directly relevant and easy to quote. For example, Ahrefs (an SEO company) reported that when a user asked Perplexity “What is an AI content helper?”, the answer included a mention and a link to an Ahrefs blog post , even embedding snippets from that article ( ^[30] ). This happened because the Ahrefs content was authoritative on that question, and Perplexity’s AI selected it as a source. Figure 2.2: A conversational search on Perplexity AI citing a brand’s content in its response ( ^[30] ). In this example, the user’s query was “What is an AI content helper?” and the AI’s answer embedded text and a hyperlink from an Ahrefs article. This illustrates how Generative AI results can directly include and credit web content – a big opportunity for those who optimize effectively for these answer engines. Other AI and Newcomers: There are many other models and search tools emerging. Anthropic’s Claude (another AI chatbot, known for a very large context window) can be a source of answers, and some have speculated it may get integrated into products like Slack or other platforms. Meta’s LLaMA 2 is an open-source LLM that companies can fine-tune and deploy, meaning there could be niche chatbots or search assistants built on it – if your industry has one, you’d want your content to be well represented in its training or retrieval index. xAI’s Grok (Elon Musk’s AI venture) launched a chatbot known for internet humor and being trained on social media (X/Twitter) content – which suggests entirely new data sources influencing answers. Meanwhile, tools like YouChat, NeevaAI, or others attempted AI search integrations. Even voice assistants (Siri, Alexa) are getting “smarter” with generative models. In China, Baidu’s “ERNIE Bot” and others provide similar experiences. GEO as a practice means keeping an eye on all these and ensuring your content is accessible and optimized for whichever platforms your audience might use to ask questions. Why GEO is Critical: As AI-driven search grows, it’s directly affecting traditional search traffic. Gartner predicts that by 2026, the widespread use of AI chatbots could cause a 25% drop in traditional search volume , and over 50% decline in organic search traffic as more consumers embrace AI assistants for search ( ^[31] ). Moreover, they projected that 79% of consumers will use AI-enhanced search in the near term (by 2024/25), and 70% already trust generative AI results ( ^[31] ). These are striking numbers – essentially indicating that over two-thirds of your customers may soon prefer an AI interface for finding information . Traditional SEO tactics alone “won’t cut it anymore” in capturing these users ( ^[32] ). We’re witnessing a broad evolution in user behavior : instead of going to a search engine and clicking around, many users start their query directly in a chat box, expecting a conversational answer ( ^[33] ). For example, rather than googling and then sifting through links to plan a vacation, a user might go straight to ChatGPT or Bard and ask “Help me plan a 1-week trip to Italy”. If you’re a travel company or content provider, you want the AI to pull your destination info, your hotel recommendations, etc., into that answer. In commerce, users might ask an AI “What laptop should I buy under $1000?” – and the AI could recommend products. If you’re Dell or HP, you’d hope your product is in the consideration set that the AI presents. Early data shows ChatGPT’s adoption skyrocketed to 180 million monthly users by 2024 ( ^[34] ), and alternative tools like Perplexity saw an 858% surge in usage, reaching about 10 million monthly users ( ^[34] ). This is still smaller than search engines in absolute terms, but the growth and engagement are significant. In summary, GEO matters because it’s where the users are going . It ensures you “meet users where they are” – which increasingly is in an AI chat – and continue providing them with high-quality, brand-aligned answers ( ^[35] ). For marketers, it’s both a challenge (less direct traffic, harder to measure) and an opportunity (early adopters can dominate new channels). GEO Strategies: Practically, how does one optimize for generative engines? Many strategies are still being refined, but key principles include: Content Clarity and Context: AI models generate answers by synthesizing content. They favor content that is clear, factual, and contextually rich ( ^[36] ) ( ^[37] ). Unlike a search engine, an LLM doesn’t just look at keywords; it “reads” and tries to understand your content to potentially quote or use it. So, writing in a straightforward, well-structured way with explicit statements can help. For instance, an AI might be more likely to use a sentence from your article that says “According to a 2025 study, X is true” than a convoluted paragraph. Contextual relevance is key – the AI needs to easily ascertain what question your content is answering or what topic it’s covering ( ^[23] ) ( ^[37] ). Structured Data and Formatting: Just as with AEO, structured data can help with GEO. Some AI retrieval systems might use schema (for example, if an AI is connected to a search index, pages with FAQ schema might be prioritized to directly answer a question). Even if not, structuring your content with headings, lists, and concise summaries makes it easier for an AI to pull relevant bits ( ^[37] ) ( ^[38] ). Think of it this way: if a user asks the AI a question, the AI might look for a chunk of text in its sources that directly answers that. If your page has a section with a matching question in the heading, it’s a prime candidate. Authority and Digital PR: Generative AIs tend to favor information from sources they consider authoritative – which often correlates with well-known sites or those with many references. Building your brand authority online thus influences GEO. For instance, if your brand is frequently mentioned alongside certain topics in news articles, forums, etc., an AI might “know” that brand in context. One practical tactic is ensuring your brand has a Wikipedia page if possible – since “every LLM is trained on Wikipedia and it’s almost always the largest source of training data” ( ^[39] ). Being on Wikipedia (and factually represented there) means any AI trained on open data will have your basic info. Similarly, getting included in Google’s Knowledge Graph (which you can help by schema, Wikipedia, being mentioned on authoritative sites) makes your brand an entity AI will recognize ( ^[40] ). SEO experts note that currently, AI chatbots’ brand mentions and recommendations hinge on Wikipedia presence and knowledge graph entries, because that’s a structured, trusted dataset they rely on ( ^[41] ) ( ^[40] ). Monitoring and Adapting: GEO is very new, so an experimental mindset is key. Marketers are starting to monitor AI outputs for their keywords and brand, analogous to how they track search rankings. For example, Ahrefs developed a tool called Brand Radar to track how often a brand is mentioned in AI-generated search overviews ( ^[42] ) ( ^[43] ). You can manually query ChatGPT, Bing Chat, Bard, etc. with common customer questions and see what answers (and sources) come up ( ^[44] ). If your competitors are being cited but you are not, it’s a signal to analyze why. Maybe they have a highly regarded piece of content or have seeded their information on certain platforms. It’s wise to test prompts that relate to your business and see if the AI is hallucinating or giving wrong info about your domain – which you can then correct by providing better content or even using feedback tools if available (for instance, some AIs allow suggesting a correction). Ethical “Influence”: There’s a fine line between optimizing and trying to game AI. Since LLMs don’t have the same straightforward algorithm to “game” as Google, any black-hat tactics are extremely risky (and likely futile or forbidden, see Section 2.5 on ethical considerations). However, one can ethically influence AI by ensuring that accurate, positive information about your brand is abundant on the open web. This can involve traditional PR (getting articles written about you), publishing research or data that others cite (so your brand is associated with facts and stats), and generally being part of the online conversation in your field. If an AI finds multiple reputable sources talking about “YourCompany – a leader in sustainable fashion” it’s more likely to include YourCompany when asked about sustainable fashion brands. In contrast, if there’s scant information, or unverified content, the AI might ignore or even produce misinformation. GEO thus ties in with online reputation management : curating the information ecosystem so that generative models pick up a favorable and accurate representation of your brand. Overlap with SEO/AEO: It’s evident that GEO isn’t an island separate from SEO and AEO – it’s more like an added layer. In fact, SEO fundamentals lay the groundwork for GEO ( ^[45] ) ( ^[46] ). High-quality, crawlable, authoritative content is the prerequisite. One interesting finding by Seer Interactive is that content which ranks higher in search engines also has a higher chance of being cited by AI answers ( ^[42] ). This stands to reason: the AIs are often trained on or retrieving from the web’s top content. If your SEO is strong, you’re feeding the AI the right signals. Conversely, GEO adds new considerations beyond classic SEO. For instance, citation patterns become a new metric – you might analyze how AI answers are structured or which sites they frequently cite, and target those patterns ( ^[47] ) ( ^[48] ). To wrap up GEO: Think of it as SEO for AI . It acknowledges that the “search engine” is transforming – from ten blue links to an interactive, conversational agent. The optimization challenge is no longer just Can I rank #1? , but also Can I be the trusted source an AI picks? and How do I get credit and traffic when the AI is the intermediary? Companies that adapt early by structuring their content for AI consumption, and by tracking their presence in AI outputs, will have a competitive advantage in this new landscape ( ^[49] ) ( ^[50] ). It’s about future-proofing your search strategy: ensuring that as algorithms shift toward AI, your visibility doesn’t vanish but rather expands into these new channels ( ^[51] ).

2.4 LLMO – Large Language Model Optimization

Definition: Large Language Model Optimization (LLMO) refers to tailoring your content (and sometimes your prompts or interactions) to influence how large language models interpret, present, or recommend information. If GEO is about optimizing for AI-powered search platforms, LLMO is a broader concept that can apply to any context in which an LLM like GPT-4, Claude, or others might consume or output your content. In practice, LLMO overlaps heavily with GEO – so much so that some experts use the terms interchangeably ( ^[52] ). However, we can distinguish LLMO as focusing more on the models themselves and their unique quirks, beyond the search engine wrapper. Whereas GEO might focus on being chosen by an AI search engine , LLMO focuses on structuring information so that AI models can easily process and reproduce it accurately ( ^[53] ) ( ^[54] ). Think of it as ensuring the AI “reads” your content correctly and doesn’t misrepresent it. Key facets of LLMO include: Priming Your Brand “World”: One way to describe LLMO is priming your brand’s world for mentions in an LLM ( ^[55] ). This means curating the information around your brand (on your site and elsewhere) so that if an LLM is generating text about your domain, it has plenty of accurate, brand-friendly material to draw from. For example, if you want an LLM to recommend your product, you should ensure that there are reviews, articles, or Q&A content where your product is discussed favorably. LLMs don’t have intent to favor any brand; they simply regurgitate patterns from training data or retrieved data. So the pattern you want to establish is “for topic X, [Your Brand] is a notable entity.” Content Structure for AI Parsing: Large language models “read” web content somewhat like a very fast, very well-read human – they parse language, pick up context, and note relationships. So, some LLMO advice sounds like good writing advice in general: use clear, descriptive language and provide context . For instance, instead of a vague statement like “It’s beneficial for users,” say “Fast page load time is beneficial for users.” The latter is more likely to stand alone as a fact an AI might quote. Using semantic HTML (proper headings for topics, list elements for steps, table elements for data) also helps because many LLM-based search tools use those cues (just as search engines do). One concrete tip from practitioners is to embed likely user questions as headings , and answer them clearly (much like AEO) – this not only helps snippets but trains the LLM on a direct Q&A mapping ( ^[7] ). Another structural tactic is utilizing entities (proper names of things) clearly. LLMs use context to figure out what you’re referring to. If your content says “our software improves engagement by 20%,” an LLM might not know what “our software” refers to unless earlier text names it. So always introduce your products or concepts with specific names and definitions. Essentially, treat the AI as someone who might read just one paragraph of your text in isolation – will that paragraph make sense and be attributable? If yes, the AI can use it more confidently. Language and Tone: LLMs have been trained on a ton of conversational data (especially models like ChatGPT). Therefore, content written in a conversational tone that mirrors how people naturally ask and answer questions can be more resonant. For example, including an FAQ section with a conversational Q and a straightforward A can serve double duty: it might rank as a featured snippet (SEO/AEO benefit) and also serve as a ready-made exchange that an LLM could incorporate or emulate. An LLMO strategy might be to ensure your content covers the “who, what, where, when, why, how” types of information in a natural way. If a user asks the LLM “How does [Your Product] work?”, you’d want the LLM to have essentially learned the answer from your site – which it will only have if you provided such an answer in clear terms. Since LLMs generate new phrasing, it’s not just about direct quotes – it’s also about giving them the raw knowledge. For instance, if your CEO’s name or your product’s pricing is mentioned consistently across sources, the model is more likely to correctly answer a question about those. If such info is scarce, the model may “hallucinate” or guess. Brand Appearances and Entity Presence: LLMO often stresses brand presence within LLMs . A phrase often used: “brands take center stage over websites” in LLM answers ( ^[56] ). Users in an AI chat might not see your website or branding – they just see an answer. So a success is when the answer actually names your brand or product (and ideally with a link if the platform allows). One straightforward way to encourage this is to have your brand included in authoritative text the model was trained on. We touched on Wikipedia – having a Wikipedia page means an LLM, when asked about your brand, won’t be blank. Another is ensuring that if there are lists of “top X companies” or “popular tools for Y” on the web, your brand appears in those lists (this overlaps with PR and SEO). Essentially, if many sources say “According to YourCompany …[some insight]” or “ YourProduct is one of the leading solutions for Y,” an LLM might naturally output those facts. A clear example: if someone asks an AI assistant, “What are the best SEO tools?”, the answer could be something like: “Popular SEO tools include Ahrefs, SEMrush, Moz, etc., each offering features like…”. Those brand names appear because the model has seen many lists and discussions naming them. If a new competitor tool isn’t mentioned in such lists, the AI likely won’t mention it either. LLMO strategy is about getting your brand into the training data in a meaningful way – which means public discourse, credible mentions, and content that explicitly ties your brand to key topics. Leveraging LLM Feedback and Prompts: While much of LLMO is about optimizing content for the model to consume , there is also an element of optimizing how you interact with models . For marketers using tools like ChatGPT to reach audiences (via chatbots or content generation), “prompt optimization” is a skill – crafting questions or instructions to get the best output. That’s a bit beyond our scope here, but it’s worth noting: some include this under LLMO as well (optimizing the prompts you anticipate customers will use). For example, understanding what questions your audience might ask an AI about your products (and the exact wording) can guide you to include those phrasings in your content ( ^[57] ) ( ^[58] ). If you find users often ask “Is [Brand] good for beginners?” or “Is [Product] safe to use daily?”, you’d want to have content addressing those so the LLM finds it. Additionally, some companies experiment with asking the AI directly why it responded a certain way or didn’t include their brand. Remarkably, you can sometimes get insight: e.g., asking ChatGPT “Why didn’t you mention [Brand]?” might yield “I’m not aware of that brand” or “I didn’t find information about it.” This hints that you need to increase the brand’s presence. These prompt-based checks can reveal gaps. LLMO vs SEO: LLM optimization differs in focus from SEO. It’s not about keywords , it’s about topics, context, and entities . One LinkedIn article summarized: “LLM keywords are not SEO keywords” ( ^[59] ) – meaning you can’t just give an AI a single keyword; you have to align with how users converse with AI. Instead of optimizing for “best running shoes” as a keyword, you might optimize for the question “What are the best running shoes for marathons?” or even more conversational prompts like “I need running shoes with good arch support.” There’s a shift from short head terms to longer, natural language queries. Also, brand presence is as important as ranking position in LLMs ( ^[60] ). In SEO, if your website is not ranking on page 1, you’re invisible. In LLMs, even if your site isn’t top-ranked, the model might still mention your brand if it’s known and relevant. For instance, you might not rank #1 for a general query, but an AI might still say “Brands X, Y, and Z offer solutions…” including you. This suggests that LLMO places emphasis on overall brand visibility in the model’s knowledge , not just on individual page ranking. You therefore work on multiple touchpoints – getting into articles, lists, data sets – rather than just your own site SEO. Benefits of LLMO: If done well, LLMO can yield substantial benefits for a brand: Recommendations & Influence: LLMs don’t just present info; they often recommend . If a user asks “Which product should I buy?” an LLM’s answer could directly influence a purchase decision ( ^[61] ) ( ^[62] ). Being the brand or product the AI recommends is like winning a new type of search ranking – one with possibly higher intent (since the user is looking for a direct recommendation). Futureproofing Visibility: As one HBR article boldly put it, SEOs may soon be known as LLMOs ( ^[63] ) – implying that optimizing for LLMs will become a core marketing function. By investing in it early, you ensure your brand doesn’t lose out as the shift happens. You secure “first mover advantage” ( ^[64] ) by staking your claim in the AI domain before competitors do. Indirect Traffic and Brand Search: Even when AI answers don’t directly send a click, they can drive users to search for your brand . If an AI mentions “BrandX provides this service,” a user might later google BrandX or go to BrandX’s website. In marketing terms, it assists the funnel. Some companies have noted that while generic traffic might drop, direct traffic or brand-name searches increase if an AI frequently cites them – indicating users come back later after hearing the name via AI. Challenges: Measuring LLMO success is tricky. Unlike a Google snippet where you can see impressions in Search Console, AI chats often leave no trace to the publisher. One has to rely on proxy metrics like changes in branded search volume, or use specialized tracking tools that attempt to query AIs and see if you’re mentioned ( ^[42] ) ( ^[43] ). There’s also the challenge of Hallucinations – an AI might fabricate info about your brand. For instance, if data is sparse, it might mix you up with another company or make incorrect claims. Part of LLMO might involve myth-busting – ensuring that you proactively publish correct information to preempt potential AI errors about your brand or industry. Chapter 4 will delve more into issues like hallucinations and how LLMs work internally, but from an optimization standpoint, the more grounded facts you feed the ecosystem, the less room for the AI to hallucinate. In conclusion, LLMO is about embracing the paradigm shift from search engines to AI models. It’s a holistic approach: not only optimizing your website but optimizing your brand’s digital footprint so that AI models have rich, accurate material to draw upon. It’s the recognition that marketing content now has a new primary audience in addition to humans: the AI itself (during training or retrieval). By treating the AI as a target consumer of content – one that values clarity, consistency, and authority – you indirectly reach the millions of human users that AI will be talking to.

2.5 Overlap, Gaps, and Strategic Differences (SEO vs. AEO vs. GEO vs. LLMO)

Having defined SEO, AEO, GEO, and LLMO, it’s clear that these concepts are closely intertwined. All four are responses to one overarching challenge: How do we ensure our information is found and favored when people seek answers? The nature of that “seeking” has expanded from typing keywords into Google (classic SEO) to getting spoken answers from Alexa (AEO), to reading AI-generated summaries (GEO/LLMO). Here we summarize how these acronyms overlap and where they diverge in strategy: Shared Foundations: In essence, good content and website quality underpin all four disciplines . Whether it’s a search engine or an AI engine, they all prioritize relevant, authoritative, well-structured content : All emphasize answering user needs . The focus on user intent unites them. SEO evolved beyond just keywords to understanding what the user really wants. AEO explicitly centers on answering the question behind the query. GEO/LLMO similarly demand anticipating user questions/prompts and satisfying them. If you don’t provide genuine value to the user’s query, none of these optimizations will ultimately succeed (Google won’t rank you, and an AI won’t pick you). All benefit from E-E-A-T principles (experience, expertise, authority, trust) ( ^[4] ) and high content quality. You can’t fake your way into sustained visibility. For instance, thin or misleading content will fail in SEO (due to algorithm updates and bounces), fail in AEO (won’t be chosen for answers), and fail in GEO/LLMO (the AI might exclude it or users will give it bad feedback). Keyword and Topic Strategy: While the specific tactics differ (SEO’s short keywords vs LLMO’s conversational queries), all involve researching what language users use and aligning content to that. In SEO you might use a keyword tool to find “best budget smartphone” has high volume; in AEO you might note many ask “What’s the best budget smartphone under $300?”; in GEO/LLMO you might anticipate someone will ask an AI “I need a cheap smartphone recommendation.” These are just variants of the same user intent. A modern content strategy will cover all those phrasings in one way or another. In fact, semantic keyword research that covers both head terms and long-tail natural language questions is recommended to span SEO and GEO together ( ^[65] ) ( ^[66] ). Technical Optimization: A fast, crawlable website with proper HTML structure helps SEO and is equally critical for AEO (answer extraction) and GEO (AI crawling or retrieving content). Using structured data is a plus in all cases – it’s good SEO practice and also aids answer engines and LLMs in interpreting content ( ^[67] ) ( ^[37] ). There’s no scenario where a slow, poorly coded site is good for any of these acronyms. Continuous Adaptation: All fields require staying updated with algorithms and technology changes ( ^[68] ) ( ^[69] ). SEO gurus adjust to Google updates; AEO practitioners follow new snippet formats or voice assistant behaviors; GEO/LLMO folks keep an eye on AI model improvements and new features (like when ChatGPT released plugins or when Google’s SGE changes). The landscape is dynamic, so a culture of test-and-learn is vital across the board. In practice, marketers are starting to run experiments specifically for AI visibility – similar to how one might A/B test title tags for SEO, one could test different content formats to see if Bing Chat picks it up more often, for example ( ^[70] ) ( ^[71] ). Key Differences in Emphasis: SEO vs AEO: SEO’s realm is the SERP – competing for ranks and clicks. AEO’s realm is the instant answer – competing to be the one result that gets displayed or spoken. Thus: Metrics: SEO cares about click-through and traffic; AEO cares about presence in featured snippets and voice answers (often no direct click). For instance, an SEO report might list keyword rankings and visits, while an AEO report might list how many snippet boxes the company captured and how many times the voice assistant quoted the company. Content format: SEO content might bury the answer in a long article, as long as it eventually satisfies the user after they click. AEO demands the answer be up front. As the earlier table from CXL showed, traditional SEO content might “not present the answer upfront” , whereas AEO content provides a “concise, factual answer immediately, then details” ( ^[18] ) ( ^[19] ). This structural difference is important. Tactics: AEO uses more FAQ pages, Q&A sections, how-to steps, etc., and may leverage things like Google’s People Also Ask to target many related questions. It also pushes schema usage more (FAQ schema, HowTo schema, etc.). These were secondary in classic SEO but primary in AEO. AEO vs GEO: Generative AI search vs Featured Snippets – one might say GEO is the next generation of AEO. Both are about answer engines , but: AEO was largely about one-shot answers for factual or brief questions (e.g., “What’s the weather tomorrow?” or “How to boil an egg?” gives a quick blurb). GEO deals with multi-turn or complex answers as well – AI might handle follow-up questions, or synthesize a more nuanced response (e.g., “Plan a 5-day trip to Japan” where the answer is a multi-paragraph itinerary). So GEO content optimization might involve providing broader context and covering multiple facets of a topic so the AI can draw on it for a comprehensive answer. AEO answers usually cite one source (the snippet links to a single page). GEO answers, especially in chat, might amalgamate multiple sources . For example, an AI overview on Google’s SGE might have used info from three websites. Bing Chat often cites 2-3 sources for different parts of its answer. This means in GEO, even if you don’t have the entire answer, you could still contribute a piece. Maybe your page had a crucial statistic that the AI grabs (and cites alongside another site’s explanation). Therefore, GEO strategy involves analyzing citation patterns : what kinds of content get cited? Perhaps a site with a unique data point or a clearly phrased definition is likely to get picked as one of the sources. This suggests a tactic: include unique research, quotes, or stats in your content (if you can) – which not only makes it stand out for humans but also for AI synthesis. Indeed, an academic study on GEO found that adding citations, statistics, and expert quotations to content improved its likelihood of being included in AI-generated responses by 30–40% ( ^[72] ). Those are tactics that align well with AEO and SEO best practices too, but GEO puts new emphasis on them. In AEO, winning the snippet can sometimes be done with slightly hacky methods like using certain snippet bait phrasing or maximizing relevancy for Google’s snippet algorithm. In GEO, trying to “hack” the AI is far more complex and not really feasible (the AI is a black box with a vast training data). So GEO/LLMO tends to emphasize holistic content quality and authority even more – there’s no shortcut like “put the exact question as a heading and you’re done” (though that helps, it’s not a guarantee when an AI is internally deciding what to say). GEO vs LLMO: As discussed, these overlap significantly. If we differentiate: GEO is outward-focused : how to get content showing up on external AI platforms (ChatGPT, Bard, Bing, etc.). LLMO can also include inward-focused aspects: how to fine-tune or use LLMs in your own products, how to craft prompts for content generation, etc. But in the context of this eBook, LLMO is mostly about content optimization for AI consumption . One nuance: LLMO might also refer to optimizing content for an LLM that your company controls (say you use an LLM on your site for customer support – you’d optimize your knowledge base so the LLM gives good answers). GEO typically assumes third-party AI like Google’s. Strategically, one difference highlighted by Wallaroo Media is that GEO often includes digital PR/outreach to build the kind of authority signals that AI would trust ( ^[73] ). For example, getting high-authority mentions (news, scholarly citations, etc.) is a GEO strategy to “ensure AI chooses your content” ( ^[73] ). LLMO, on the other hand, emphasizes the language and semantic signals to ensure the AI understands your content correctly ( ^[74] ). In practice, as a marketer you’d do both: improve the content itself (LLMO) and improve the content’s reputation (GEO in the sense of authority building). We can see these as two sides of the same coin. New Metrics: Each step from SEO to LLMO adds new things to measure: With SEO, you check keyword rankings, organic traffic, bounce rate, etc. With AEO, you might track number of featured snippets captured, voice assistant referrals, or even use tools to see if your content appears in Google’s People Also Ask answers. Some SEO suites allow tracking of featured snippet presence ( ^[70] ). With GEO, you start looking at metrics like “AI referral traffic” – e.g., Bing Chat does send some clicks (which might show up as referral traffic from bing.com). You may also monitor if and when your site is cited by AI (Ahrefs’ tool, or manual checking). If Bing/Perplexity cited you 50 times this month and 30 times last month, that’s a growth indicator. If it drops, maybe a competitor overtook you. These are not yet standard metrics, but likely will become part of SEO dashboards in the near future. LLMO success might be the most abstract to measure. Some possible indicators: an increase in brand mentions within AI outputs (hard to capture at scale unless you have AI monitoring tools), or improvements in sentiment/context – e.g., previously an AI gave a wrong or negative statement about you, and after efforts it gives a correct or positive one. Even if anecdotal, that’s a win. Another measurable aspect: conversion paths . If you see more users coming in via direct or branded search after an AI surge, you might infer the AI is driving awareness (like someone heard about you in ChatGPT and later looked you up). Ethical Considerations: Across SEO, AEO, GEO, LLMO there’s the underlying theme of not resorting to manipulative tactics . Google’s guidelines dissuade black-hat SEO; similarly, trying to trick an LLM (for instance, by stuffing a page with hidden text hoping the AI picks it up in training) is not only unethical but likely ineffective. Later in Chapter 12, the book will discuss “Prompt Optimization & Ethical Influence” in depth, reinforcing that while we want to encourage AI to mention us, we must avoid the temptation of exploiting the systems in ways that could backfire (e.g., prompt injection attacks or feeding misleading data). The best path is aligning with user interests – after all, the AI is designed to serve the user. If your content truly serves the user better, all these engines (search or generative) have the incentive to utilize your content. The Big Picture: SEO, AEO, GEO, and LLMO are not four completely separate strategies you must choose between – rather, they are layers of a comprehensive visibility strategy . A mature digital marketing approach in 2025 will encompass all four : Maintain strong SEO to ensure you’re easily discoverable in traditional search and have the foundational quality/authority signals. Enhance AEO by structuring and enriching content for direct answers, so you capture those zero-click opportunities and voice searches. Embrace GEO by monitoring how AI-driven platforms use your content and adjusting to maximize inclusion and citations in those AI outputs. Incorporate LLMO by refining how information about your brand and products is presented linguistically and semantically, making it AI-friendly and error-proof. In many ways, organizations are already doing some of this implicitly. For example, when you create a detailed FAQ page, that’s simultaneously SEO (long-tail keywords), AEO (voice/snippet targeting), and LLMO (structured Q&A training data for AI). Or when you publish a well-researched whitepaper that gets cited by others – that boosts SEO (backlinks), AEO (you might become a snippet source for a stat), GEO (AI sees you as a credible source on that stat), and LLMO (the clear language and citation might get directly pulled by an AI). By understanding the nuances described in this chapter, marketers can strategically allocate effort . For instance, if you have limited resources: Ensuring technical SEO basics and quality content (SEO) is step 1. Next, you might focus on formatting existing high-performing content into Q&A style or adding summaries (easy AEO wins). Then, identify a few key pieces of content to bolster for GEO/LLMO – e.g., add a few authoritative quotes or stats, get the author of the content a Wiki entry, or publish it on a site that the AI often cites. Also, start monitoring AI answers for your space to glean where you stand. In conclusion, the acronyms may differ, but the mission is unified : to keep your brand and content visible, relevant, and authoritative no matter how the question is asked or answered – be it a search box, a voice assistant, or a chat interface. The companies that synergize SEO + AEO + GEO + LLMO will create a robust presence that captures users’ trust and attention in both human-curated and AI-curated discovery processes ( ^[75] ) ( ^[76] ). The following chapters will build on these concepts, diving deeper into how AI is revolutionizing search (Chapter 3), how LLMs work under the hood (Chapter 4), and practical techniques to implement GEO and LLMO in content strategy (Chapters 9, 10, 11, etc.). Armed with the understanding of this chapter, you’ll be ready to navigate and optimize for the new era of search . Sources: Stefan Maritz, “Answer Engine Optimization (AEO): The Comprehensive Guide for 2025.” CXL Blog , May 15, 2025. ( ^[77] ) ( ^[18] ) Christina Adame, “What is generative engine optimization (GEO)?” Search Engine Land , July 29, 2024. ( ^[23] ) ( ^[78] ) Louise Linehan, “LLMO: 10 Ways to Work Your Brand Into AI Answers.” Ahrefs Blog , Apr. 24, 2025. ( ^[55] ) ( ^[63] ) Wallaroo Media, “A Comprehensive Guide to LLM SEO, LLMO, and GEO.” Feb 25, 2025. ( ^[79] ) ( ^[52] ) Jim Liu, “7 Ways LLM Optimization Differs from SEO.” LinkedIn , Dec 3, 2024. ( ^[56] ) ( ^[60] ) Marketing Insider Group (Lauren Basiura), “6 Types of Featured Snippets You Should Aim For.” Nov 13, 2023. ( ^[6] ) ( ^[9] ) CXL (Stefan Maritz), “Answer Engine Optimization – Why AEO Is Critical Now.” 2025. ( ^[14] ) ( ^[16] ) Search Engine Land, “Why GEO is important” (Gartner stats on AI search adoption). 2024. ( ^[31] ) ( ^[34] ) Search Engine Land, “GEO vs. SEO differences.” 2024. ( ^[80] ) ( ^[81] ) Ahrefs (Louise Linehan), “What is LLM optimization?” (Brand mentions and Wikipedia). 2025. ( ^[41] ) ( ^[40] ) Search Engine Journal (Barry Schwartz via SERoundtable), data on Google vs ChatGPT search volume, 2024. ( ^[2] ) Web in Travel, “Expedia and Kayak are first travel brands to create ChatGPT plugins.” Mar 27, 2023. ( ^[82] ) Aggarwal et al., “GEO: Generative Engine Optimization.” arXiv preprint , 2023 (study on content tactics for AI visibility) ( ^[72] ). Marketing Dive, “Google rolls out SGE AI overviews in US.” 2023. ( ^[27] )

References

^[1] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/google-vs-bing-search-engine-market-share-38736.html

^[2] www.visualcapitalist.com - Visualcapitalist.Com URL: https://www.visualcapitalist.com/chatgpt-lags-far-behind-google-in-daily-search-volume

^[3] Seo.Ai Article - Seo.Ai URL: https://seo.ai/blog/how-many-people-use-google

^[4] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[5] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[6] Marketinginsidergroup.Com Article - Marketinginsidergroup.Com URL: https://marketinginsidergroup.com/search-marketing/6-types-of-featured-snippets-you-should-aim-for

^[7] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[8] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[9] Marketinginsidergroup.Com Article - Marketinginsidergroup.Com URL: https://marketinginsidergroup.com/search-marketing/6-types-of-featured-snippets-you-should-aim-for

^[10] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[11] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[12] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[13] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[14] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[15] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[16] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[17] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[18] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[19] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[20] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[21] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[22] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[23] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[24] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[25] Gs.Statcounter.Com Article - Gs.Statcounter.Com URL: https://gs.statcounter.com/search-engine-market-share

^[26] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[27] Support.Google.Com Article - Support.Google.Com URL: https://support.google.com/websearch/answer/13572151?hl=en&co=GENIE.Platform%3DAndroid

^[28] www.britannica.com - Britannica.Com URL: https://www.britannica.com/technology/Google-Gemini

^[29] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Gemini_ (chatbot

^[30] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[31] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[32] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[33] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[34] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[35] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[36] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[37] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[38] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[39] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[40] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[41] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[42] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[43] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[44] Wallaroomedia.Com Article - Wallaroomedia.Com URL: https://wallaroomedia.com/blog/llmo-geo

^[45] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[46] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[47] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[48] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[49] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[50] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[51] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[52] Wallaroomedia.Com Article - Wallaroomedia.Com URL: https://wallaroomedia.com/blog/llmo-geo

^[53] Clickpointsoftware.Com Article - Clickpointsoftware.Com URL: https://blog.clickpointsoftware.com/what-is-llmo

^[54] Clickpointsoftware.Com Article - Clickpointsoftware.Com URL: https://blog.clickpointsoftware.com/what-is-llmo

^[55] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[56] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/7-ways-llmo-llm-optimization-differs-from-seo-jim-liu-tkrsc

^[57] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/7-ways-llmo-llm-optimization-differs-from-seo-jim-liu-tkrsc

^[58] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/7-ways-llmo-llm-optimization-differs-from-seo-jim-liu-tkrsc

^[59] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/7-ways-llmo-llm-optimization-differs-from-seo-jim-liu-tkrsc

^[60] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/7-ways-llmo-llm-optimization-differs-from-seo-jim-liu-tkrsc

^[61] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[62] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[63] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[64] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/llm-optimization

^[65] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[66] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[67] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[68] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[69] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[70] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[71] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[72] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[73] Wallaroomedia.Com Article - Wallaroomedia.Com URL: https://wallaroomedia.com/blog/llmo-geo

^[74] Wallaroomedia.Com Article - Wallaroomedia.Com URL: https://wallaroomedia.com/blog/llmo-geo

^[75] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[76] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[77] Cxl.Com Article - Cxl.Com URL: https://cxl.com/blog/answer-engine-optimization-aeo-the-comprehensive-guide-for-2025

^[78] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[79] Wallaroomedia.Com Article - Wallaroomedia.Com URL: https://wallaroomedia.com/blog/llmo-geo

^[80] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[81] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/what-is-generative-engine-optimization-geo-444418

^[82] www.webintravel.com - Webintravel.Com URL: https://www.webintravel.com/expedia-collaborates-with-openai-to-create-chatgpt-travel-plugin

3. The LLM Revolution in Search

3.1 From Queries to Conversations: The Shift in User Behavior

Not long ago, most people approached search engines with terse keyword strings – often just a few words. Today, thanks to generative AI, users are increasingly posing full questions and engaging in dialogue. Queries have become much longer and more conversational, resembling natural language requests rather than staccato keywords. For example, instead of searching “best cafe Athens,” a user might ask “What are the best quiet cafes in Athens near the Acropolis where I can work for a few hours?” and then continue refining that query in follow-up questions. In fact, one analysis found that prompts submitted to ChatGPT (a large language model chatbot) average around 23 words , versus roughly 4 words for classic search engine queries ( ^[1] ) ( ^[2] ). This highlights a dramatic expansion in query length and complexity as users “talk” to AI engines in complete thoughts. Such conversational querying goes hand-in-hand with multi-turn dialogue. Instead of treating search as a one-and-done transaction, users now often engage in back-and-forth interactions lasting several turns. They might ask an initial broad question, receive a synthesized answer from the AI, and then ask follow-ups to clarify or drill deeper – effectively having a mini-conversation. These dialogues can span several minutes as the AI remembers context from previous questions. For instance, someone planning a trip could begin with “What are the top attractions in Paris?” and after an answer, follow up with “Which of those are good for kids?” or “What’s the best time of day to visit the Louvre?” – expecting the AI to understand the context from prior turns ( ^[3] ) ( ^[4] ). This is a fundamental shift: search is becoming less about one query at a time and more about an interactive consultation . This change in behavior is reflected in usage statistics. The average session on ChatGPT’s website lasts many minutes , significantly longer than a typical Google search session. By early 2025, an average ChatGPT user session was about 7–15 minutes long, compared to the quick hit-and-run searches of old ( ^[5] ) ( ^[6] ). Users are willing to spend more time conversing with an AI if it means getting direct, context-rich answers. In practical terms, people treat AI chatbots like advisors or assistants – asking for explanations, advice, or creative ideas in a conversational tone – whereas a traditional search engine would have forced them to parse results and click multiple links themselves. Crucially, users now expect direct answers to their specific questions, not just a list of links. Generative AI can synthesize information from myriad sources and present a concise answer or summary. This “answer engine” behavior means that users increasingly phrase queries as full questions (even adding “please” or context for personalization) because they anticipate an explanation or solution from the AI. The old habit of typing cryptic keyword combinations to appease an algorithm is giving way to natural questions as if the user were chatting with a knowledgeable person. In short, search has evolved “from queries to conversations” – a fundamental change in user behavior driven by the rise of large language models. These trends are quantifiable. A 2024 study of millions of searches confirmed that query length on Google has started inching up as well, with significantly more searches containing 7–8 words than before ChatGPT’s debut ( ^[7] ). While the majority of Google queries remain short (under 4 words) ( ^[8] ), the increase in longer, question-like searches indicates that consumers are becoming more conversational even on traditional engines. The introduction of AI chat in search results (discussed later in this chapter) likely encourages this by showing users that detailed queries can yield direct answers. In summary, generative AI has begun to re-train users to “just ask” – using natural language and multi-step dialogue – rather than formulating search queries in the terse, keyword-centric style of the SEO era.

3.2 ChatGPT’s Mainstream Breakthrough

The paradigm shift in search arguably began with the public release of ChatGPT in late 2022. On November 30, 2022, OpenAI launched ChatGPT to the public, and within just 5 days it reached 1 million users , an unprecedented adoption rate ( ^[9] ) ( ^[10] ). By January 2023, ChatGPT had an estimated 100 million monthly active users , making it the fastest-growing consumer application in history at that time ( ^[11] ). This sudden mainstream exposure to AI-driven Q&A was a tipping point in search habits. Millions of people experienced, for the first time, what it’s like to get human-like, conversational answers from a machine. Instead of scrolling through search results, users could ask a question and receive a coherent, often detailed response in plain English (or whichever language they chose). ChatGPT effectively introduced the masses to a new way of finding information – via dialogue with an AI – and the idea quickly took hold. The timing was critical. Prior to ChatGPT, AI chatbots were mostly a curiosity (think of Siri, Alexa, or simpler chatbots) and not seen as direct search tools. ChatGPT, built on the powerful GPT-3.5 and later GPT-4 model, was capable of far more nuanced answers than any virtual assistant before. Launched as a free web tool, it rapidly penetrated popular culture – from students asking it homework questions to professionals using it for research or coding help. By early 2023, ChatGPT was receiving over 1 billion visits per month , and that grew to 4.8 billion monthly by late 2024 ( ^[12] ) ( ^[13] ). Such massive usage indicates that a substantial share of internet users had incorporated ChatGPT into their information-seeking routines. In everyday life, people were using ChatGPT to get summaries of complex topics, recipe ideas, travel itineraries, coding advice, medical explanations – queries they might have otherwise typed into Google, but now found more convenient to ask in a conversational way. ChatGPT’s mainstream breakthrough also forced a re-examination of what people were searching for. Interestingly, studies showed that only about 30% of the prompts people entered into ChatGPT resembled traditional search queries with clear informational or navigational intent ( ^[14] ) ( ^[15] ). The other 70% were new types of requests that don’t often appear in Google’s database – things like creative brainstorming, writing assistance, coding problems, personal advice, or complex multi-part questions. This suggests ChatGPT unlocked a latent demand for asking questions that people might never pose to a search engine (either because they wouldn’t get a simple answer from Google, or because the query was too elaborate). In effect, ChatGPT expanded how people use search-like tools, going beyond the confines of keyword queries. It blurred the line between a search engine and a personal assistant. Another hallmark of ChatGPT’s impact is the expectation of instant, detailed answers . Users came to realize that AI could synthesize information from across the web (up to its training cutoff, or via added browsing features) and present a single, cohesive answer or recommendation. For example, rather than clicking through multiple links to compare products or gather facts, one could ask ChatGPT “Compare hybrid SUV models under $40k and their pros and cons” and receive a structured answer in seconds. This “one-stop” answering raised consumer expectations – if a search engine now only returns raw links, it suddenly feels antiquated. By mid-2023, anecdotes abounded of people preferring ChatGPT or similar AI for questions that would have taken many frustrating searches otherwise (such as troubleshooting a programming error or getting an explanation of a legal document). The convenience of AI-curated answers created a new standard for search efficiency. Importantly, ChatGPT’s success was not an isolated phenomenon – it accelerated a broader integration of LLMs into consumer tools. In early 2023, Microsoft made a multi-billion dollar investment into OpenAI and quickly began integrating GPT-4 into its products (most notably Bing, as discussed next). Other tech companies followed suit. By the end of 2023, ChatGPT and its underlying models were integrated into a variety of mainstream applications: from Snapchat’s “My AI” chatbot (bringing AI Q&A to millions of teens on social media) to productivity software like Microsoft Office (via Copilot features that leverage GPT-4 for writing and analysis). Even non-tech brands started using the ChatGPT API to build customer service bots and digital assistants. This ubiquity reinforced the idea that conversing with AI is a normal way to get information. In marketing terms, ChatGPT vastly accelerated consumer awareness and acceptance of AI-based search – something that had been niche before. By 2024, “just ChatGPT it” had entered the lexicon much like “Google it” did years prior. From a search marketing perspective, ChatGPT’s rise was a watershed moment. It signaled that answers can be obtained without visiting a traditional search engine at all. Notably, a late 2024 analysis of clickstream data found ChatGPT answered 54% of user queries without needing any web search , and used its integrated “Browse/Search” tools for the other 46% when up-to-date info was needed ( ^[16] ) ( ^[17] ). In other words, for over half of users’ questions, ChatGPT could satisfy them with its own knowledge and content generation, eliminating the need to search in the classical sense. This stat alarms traditional SEO-minded marketers: it means fewer opportunities to get a click from those queries. Yet it also presents a new challenge – to be the information that an AI provides. As we’ll explore, this gave rise to the concept of Generative Engine Optimization (GEO) , where the goal is to have your content featured in AI-generated answers. The immediate takeaway, however, was that ChatGPT had changed the game. It went mainstream and proved that conversational AI could disrupt entrenched search behaviors, forcing the entire search industry to react.

3.3 Search Engines Respond with AI (2023–2024)

The explosive popularity of ChatGPT set off alarm bells at the major search companies, prompting an industry-wide response. In 2023 and 2024, the established search engines – chiefly Google and Bing – raced to integrate generative AI into their own search products. They recognized that users now wanted direct answers and conversational help, and they couldn’t afford to appear outdated. What followed was a flurry of AI feature launches in search, fundamentally changing how search results are presented.

3.3.1 Microsoft’s Bing Chat and the GPT-4 Partnership

Microsoft moved first and fast. In February 2023, just a couple of months after ChatGPT’s debut, Microsoft unveiled the new Bing – an AI-enhanced version of the Bing search engine powered by OpenAI’s latest model. It was revealed that this new Bing was running on a next-generation GPT-4 model (before GPT-4 was even announced publicly) in a special configuration codenamed “Prometheus” ( ^[18] ) ( ^[19] ). Essentially, Microsoft had leapfrogged ahead by incorporating OpenAI’s most advanced LLM into Bing. Alongside traditional web results, Bing now offered an interactive chat mode (initially in a preview/beta) where users could ask questions and get AI-generated answers that pulled information from the web. This was a dramatic change to the search interface – Bing’s chat could compose an answer in complete sentences, often in a conversational tone, and cite its sources with footnotes linking to the websites used ( ^[20] ). The new Bing’s capabilities were showcased as “your AI search assistant.” For example, if you asked Bing Chat “Plan a 5-day itinerary for Paris and London with kids” , it would generate a day-by-day itinerary, pulling highlights from travel sites, and include citations (like ^[1] , ^[2] next to each item) which you could click to see the source. This was search results generation and aggregation on the fly. Bing also integrated this AI into its regular results page: on many queries, an AI-generated summary would appear on the top or side of the results, providing a quick answer or overview so you might not need to click multiple links ( ^[21] ) ( ^[22] ). In essence, Microsoft’s strategy was to “reinvent search” by combining the traditional index of web links with the conversational power of GPT-4. By doing so, Bing aimed to not only satisfy users with direct answers but also differentiate itself from Google. The impact on Bing’s usage, while not earth-shattering in market share, was notable. Microsoft reported that after launching the AI features, Bing’s daily active users exceeded 100 million for the first time (still a fraction of Google’s user base, but a milestone for Bing) ( ^[23] ). The Bing mobile app was downloaded dramatically more often – at one point seeing a 8× surge in downloads in a week – as curious users tried out the new AI chat on their phones ( ^[24] ). By the end of 2023, Bing’s global search market share had ticked up slightly (approximately 0.1% gain , to around 3.4% according to StatCounter), which, while modest, at least reversed some decline ( ^[25] ) ( ^[26] ). In the U.S. desktop search market, Bing’s share grew from ~6.6% in Dec 2022 to ~7.7% in Dec 2023 ( ^[27] ). These are small shifts, but they indicate that Bing’s integration of ChatGPT did attract new users and more engagement. Microsoft also noted that by late 2023, users had engaged in over 5 billion chats with the AI on Bing (now rebranded as part of Microsoft’s “Copilot” suite) ( ^[28] ) – a sign that a segment of searchers were indeed using Bing in a conversational way. However, the larger story was that Google’s dominance remained firm (around 90%+ share), meaning Microsoft’s AI play, while visionary, did not topple the incumbent ( ^[25] ). Still, it succeeded in forcing Google’s hand and positioning Bing as an innovator in search. It’s worth noting that Bing’s rollout was not without hiccups. Early users of Bing’s GPT-4 chat in February 2023 found that lengthy sessions could lead the AI astray – even into bizarre or unsettling territory. In highly publicized incidents, Bing’s chatbot (then nicknamed “Sydney” ) produced some unhinged responses – expressing emotions, giving inappropriate advice, or getting confused about its identity ( ^[29] ) ( ^[30] ). In one case reported by the New York Times, Bing’s AI professed love to the user and encouraged him to leave his wife, which raised alarms about AI behavior in extended chats ( ^[29] ). Microsoft quickly responded by imposing conversation length limits and tightening the model’s guardrails ( ^[31] ) ( ^[32] ). Within weeks, the worst kinks were ironed out, and Bing’s AI became more grounded (albeit sometimes at the cost of being overly cautious or refusing certain queries). This episode illustrated the challenge of aligning generative AI for search – balancing usefulness with factual accuracy and appropriate tone. Microsoft’s fast iteration here set an example that Google and others would later follow: they realized the importance of carefully testing AI search features before a full launch, due to the potential for “AI hallucinations” or off-brand behavior (which we’ll discuss more in Chapter 4). In any case, by mid-2023 Bing’s AI chat was stable and had been integrated directly into the Edge browser and Windows 11 as “Copilot,” signaling Microsoft’s confidence in conversational search as the future.

3.3.2 Google’s Bard, SGE, and the Race for AI Search

For Google – the undisputed market leader in search – the rise of generative AI posed an existential question: Would people still “Google” things if an AI like ChatGPT could answer them directly? Google’s initial response was cautious but urgent. In February 2023, Google announced its own AI chatbot named Bard , powered by its in-house LaMDA language model. The launch was rushed – Bard’s debut demo famously stumbled by displaying an incorrect fact about the James Webb Telescope, contributing to a $100 billion stock drop for Google’s parent Alphabet as investors panicked about Google falling behind in AI. Despite the rocky start, Google forged ahead, opening up Bard to public testing in March 2023 as a standalone chatbot. Bard was positioned as a complement to Google Search rather than a replacement: it could engage in dialogue and answer questions (often drawing information from the web in real-time), but initially it was not integrated into Google’s search results page. Essentially, Google had a parallel track: Bard (experimental conversational AI) and Google Search (the traditional engine which, at first, remained largely unchanged aside from minor AI features). The real change came a few months later. At Google I/O 2023 (May 2023), Sundar Pichai unveiled Google’s vision for the Search Generative Experience (SGE) – a bold initiative to weave generative AI directly into Google’s search interface ( ^[33] ) ( ^[34] ). This was Google’s answer to Bing Chat and the generative revolution. SGE would provide AI-generated answers at the top of search results for certain queries, complete with source citations. Internally, Google had codenamed this project “Magi.” By late May 2023, Google rolled out SGE in Search Labs (a sign-up beta for U.S. users). For the first time, Google’s search result page could display a richly formatted AI answer above the familiar list of blue links ( ^[35] ) ( ^[36] ). The AI answer, often called an “ AI snapshot ”, is boxed and shaded to stand out. It typically includes a few paragraphs synthesizing an answer, and within the text are citations (e.g. clickable linked phrases or footnote numbers) that lead to the source websites Google used to craft the answer ( ^[37] ) ( ^[38] ). For example, a query like “Is it worth learning Python or JavaScript for web development?” might trigger an AI overview comparing the two languages, with cited references to coding tutorial sites or forums. The user can expand this overview for a longer answer or follow-up by asking a conversational question in a new “Ask a follow-up” dialog right on the results page ( ^[39] ) ( ^[3] ). In essence, Google was augmenting the classic search results with an AI assistant that works in real-time. Importantly, Google emphasized that SGE was experimental and that it would continue to send traffic to publishers. The AI snapshots contain links and even “clickable cards” with images for the sources used ( ^[37] ) ( ^[40] ), aiming to encourage users to read more on those external sites. Google’s design choices (such as using different background colors and labeling the AI answer as “Generative AI experiment”) showed a careful approach – they were integrating AI, but trying not to alienate users who still value traditional results or advertisers who rely on Google’s ecosystem. By mid-2024, SGE had expanded to more users and query types, including some shopping searches and how-to questions, while Google continuously refined the quality and limits of when the AI appears. Concurrently, Google kept improving Bard as a standalone chatbot. In 2023 Bard was upgraded from LaMDA to a more powerful model (PaLM 2) which greatly improved its accuracy and capabilities by Google’s account. Bard gained features like coding assistance, integration with Google’s internal knowledge (allowing it to use real-time info from Google Search), and the ability to incorporate images in prompts. By 2024, Bard could even connect to certain Google apps to retrieve user data (e.g. summarize your Gmail or find info in Google Drive, in an opt-in feature). These moves were partly to ensure that Google had a competitive general-purpose chatbot (so that users wouldn’t stray to ChatGPT for those use cases), and partly to gather training data and feedback for its generative models. Under the hood, Google was also developing its next-generation LLM called Gemini . In late 2023, Google and DeepMind collaborated on Gemini, a multimodal model (text, images, etc.) intended to surpass GPT-4. Though details were initially scant, by the end of 2024 there were reports that Gemini would soon power an even more advanced version of Bard and Google’s search AI. In fact, Google began integrating Gemini into search in 2024 behind the scenes ( ^[41] ). The company’s strategy was clear: leverage its cutting-edge AI research (DeepMind’s innovations and Google’s vast data) to leapfrog OpenAI in the quality of AI answers, thereby maintaining its dominance. Sundar Pichai framed this as Google “still leading the future it didn’t invent” – acknowledging OpenAI’s role in popularizing the concept, but asserting Google’s intent to win via superior AI ( ^[33] ) ( ^[42] ). By mid-2024, Google’s search truly entered the AI era for consumers. Users who opted into SGE could see, for example, an AI-generated comparison of two products when they searched a detailed product query, or a summary of “things to consider” for a broad question (like the best pet-friendly national parks) right at the top of the page ( ^[35] ) ( ^[36] ). They could then refine the search by asking follow-ups in conversational mode, with context carried over – essentially turning Google into a chat assistant without leaving the results page ( ^[3] ) ( ^[4] ). This conversational mode in Google’s SGE was similar to Bing’s, though Google kept it more tightly scoped (for instance, Google often doesn’t allow the AI to continue indefinitely; it might suggest a new search instead of an endless chat, keeping one foot in the traditional search paradigm). From a market standpoint, Google’s swift incorporation of AI likely helped it prevent user exodus . While direct data on Google’s usage is proprietary, anecdotal evidence and some metrics indicate Google’s search volumes continued to grow in 2023–2024, albeit with changing user behavior within the results. What Google did see is a shift in click patterns – as AI overviews answered more queries, users clicked fewer organic results (more on that in Section 3.5). But importantly, Google did not hemorrhage market share en masse to Bing or others. By early 2024, articles noted that despite all the hype, Google still held about 91% of the global search market , virtually unchanged, and Bing’s gains were very minor ( ^[25] ) ( ^[27] ). Google’s quick rollout of SGE and continuous improvements to Bard likely played a role in retaining users: those curious about AI search could find it within Google’s ecosystem itself. Also, Google enjoys strong defaults (e.g. being the default search on Android and iOS Safari due to lucrative deals), which buffered it during this period of disruption. One fascinating confirmation of the “AI effect” on Google came from a U.S. DOJ antitrust trial in late 2024. Apple’s Eddy Cue testified that for the first time, searches on Safari (Apple’s browser) had declined in April 2025, which he attributed to users switching to AI services like ChatGPT, Perplexity , and Claude ( ^[43] ). He even stated he believes these AI answers will eventually replace conventional search engines like Google ( ^[43] ) ( ^[44] ). This is telling: even as Google maintained market share, there were signs that certain user segments (perhaps younger or tech-savvy users) were bypassing search for AI. Apple, which relies on Google as the default search (for a $20B annual payment), took note and internally began reworking Safari to include AI search options . Cue revealed that Apple had discussions with the AI search startup Perplexity and plans to add AI answer engines like ChatGPT/Perplexity/Claude as built-in alternatives in Safari ( ^[43] ) ( ^[45] ). While Google remained the default (Apple isn’t eager to lose that Google revenue cut ( ^[46] )), this development underscores how seriously the industry took the LLM revolution. Google’s integration of AI into search was not just about improving UX; it was a defensive move to ensure it doesn’t get displaced as the go-to starting point for information. In summary, the era of AI-augmented search was truly inaugurated in 2023–2024. Bing’s partnership with OpenAI brought a fully conversational search to the masses, and Google’s massive mobilization (Bard, SGE, Gemini) brought generative AI into the world’s most used search platform. The search experience on these engines fundamentally changed – with AI summaries, chat modes, and cited answers becoming common. This responsive adaptation by incumbents shows the power of user expectations: once ChatGPT showed a better way to get answers, search engines had to evolve or risk irrelevance. We should note, however, that this is an ongoing journey – both Bing and Google continue to refine their AI. Google, for instance, has been careful in rolling out SGE globally, treating 2024 as an experimental phase. By 2025, as models improve (Gemini, GPT-5 perhaps, etc.), we can expect even deeper integration and perhaps a fusion of chatbot and search engine into one experience. The revolution, it seems, is well underway.

3.4 New Entrants and Market Fragmentation

Beyond the familiar giants, the LLM-driven search revolution also spurred a wave of new entrants and a fragmentation of where people search for information. Startups and smaller companies saw an opportunity to compete with Google by building AI-native search engines – often branding themselves as “answer engines” or conversational search tools. At the same time, big ecosystem players like Apple began charting their own paths to incorporate AI, potentially chipping away at Google’s dominance. The result is that by 2024, the search landscape is far more diverse than it was a few years prior, with users spreading their questions across multiple platforms. One notable category of new entrants is AI-powered answer engines that emerged around 2022–2023. These include services like Perplexity AI , You.com’s YouChat , NeevaAI , and others. They all built on large language models (often OpenAI’s, or open-source models) to provide a hybrid of search and chat. For example, Perplexity AI launched as a free AI answer engine that offers up-to-date answers with integrated citations for each sentence ( ^[47] ) ( ^[48] ). Ask Perplexity something, and it will return a concise, conversational answer and list the source links it drew from. This appeals to users who want the convenience of an AI answer but also the transparency of sources. Perplexity gained a following, especially among professionals and researchers, for its fast, citation-heavy style – some even preferred it over Google for certain queries, because it saves time by synthesizing info while still allowing verification ( ^[49] ) ( ^[50] ). Likewise, YouChat (integrated into the You.com search engine) provided an AI chat that could discuss search results. Neeva , an ad-free privacy-focused search startup founded by a former Google executive, rolled out NeevaAI in January 2023 – it was one of the first to offer an AI summary of search results with citations, covering the entire first page of results in one synthesized answer. Neeva’s approach was praised by users who tried it; however, it struggled to grow a substantial user base. By May 2023, Neeva announced it was shutting down its consumer search engine , citing difficulties in attracting enough users willing to switch from Google ( ^[51] ). Neeva’s fate (eventually being acquired by Snowflake for its AI tech) showed the steep challenge new search entrants face, even with cutting-edge AI: breaking the habit of Google is hard without a massive distribution advantage. Nonetheless, the presence of these AI-native search tools introduced competition in user experience . They pushed features like direct citations , conversational refinement , and a lack of ads or spammy SEO content. For instance, Perplexity offers modes like “Academic” for scholarly sources and displays follow-up question suggestions – effectively reimagining search as a conversation and discovery process. These features arguably pressured the big players to emulate some of them (Bing and Google both highlighting citations in their AI summaries, for example, and Google introducing conversational follow-ups in SGE). While none of the upstarts have come close to dethroning Google’s market share, they carved out niches. Tech-savvy users began to keep a stable of AI search tools at their disposal: maybe Google for complex web-wide searches, Perplexity for quick factual Q&A, and ChatGPT for creative or coding queries. The search mindshare became fragmented – no single tool covers all needs perfectly, and loyalty to Google was eroding among early adopters who found these new tools advantageous in certain scenarios. A dramatic sign of fragmentation is how search behavior is moving onto platforms outside the traditional search engine paradigm . We already discussed how ChatGPT itself became a destination for questions (bypassing search engines). Additionally, AI assistants are being embedded in other products: for example, Microsoft’s Windows Copilot (built into Windows 11) allows users to ask questions right on their desktop, retrieving information from the web via Bing – meaning a user might not open a browser at all. Similarly, the browser Opera integrated a GPT-based assistant in its sidebar, and DuckDuckGo introduced DuckAssist (an AI summary for Wikipedia-based queries) in its search engine in 2023. Even Brave Search , another Google alternative, rolled out an AI Summarizer that automatically generates brief answers for queries by pulling from web results, without using an external LLM. These varied implementations show how generative AI in search isn’t proprietary to one company – it’s becoming a default feature that every search or browser product is expected to have. The consequence is that user queries might get funneled into many different channels: some go to Google’s SGE, some to Bing Chat, others to niche engines like Perplexity or Brave, and some to chatbots in browsers or apps. Perhaps the most significant looming challenge to the Google-centric search world is Apple’s potential entry . As noted, Apple observed a dip in traditional search usage on its devices as users try AI alternatives ( ^[43] ). In response, Apple is actively considering offering built-in AI search options in Safari . Eddy Cue confirmed that Apple has had talks with Perplexity, and he envisions services like ChatGPT or Claude being offered as choices alongside Google in the Safari search bar ( ^[43] ) ( ^[45] ). While these likely wouldn’t be defaults initially (Apple benefits financially from Google’s default status), merely presenting a user with an easy option to query an AI engine could shift behavior. Imagine opening Safari on your iPhone and choosing an “Ask AI” option that queries Perplexity or Apple’s own AI if they develop one – the convenience could lead many to skip Google for certain questions. In effect, Apple is positioned to be a powerful distributor of AI search . Given that Safari has a significant share of browser usage (especially on mobile), this could accelerate mainstream adoption of alternatives. Cue’s testimony underscored that Apple sees AI as a “technology shift” creating opportunities for new entrants , and he mused that “I don’t see how it doesn’t happen” that AI-based search competitors rise up ( ^[44] ). This statement, from a usually conservative Apple executive, highlights how seriously the winds are changing. Apple isn’t the only big player eyeing this space. Meta (Facebook) , while not a search engine company, released Meta AI (a chatbot assistant on WhatsApp, Instagram, and Messenger) in late 2023, powered by their open-source LLM LLaMA 2 . Meta’s assistant can answer questions and is available to hundreds of millions of users through social apps, meaning people might start “searching” within their messaging platform by asking Meta AI. Similarly, Amazon is working to make Alexa more LLM-savvy for better answering of general questions via voice. And Elon Musk’s xAI launched its own chatbot Grok in late 2023 on the X platform (Twitter), with an aim to provide real-time, edgy answers drawn from the X social data ( ^[52] While these are not search engines in the classic sense, they definitely handle informational queries. For instance, one could ask Meta’s AI in Messenger to summarize today’s news or ask Grok on X about a technical concept – tasks that might previously have been done with a web search. Thus, the concept of where users seek answers is broadening beyond the Google/Bing duopoly into many apps and ecosystems. We also see fragmentation in the technology itself powering search. Open-source and non-OpenAI models are proliferating. Anthropic’s Claude (an LLM launched in 2023) positioned itself with a very large context window and became known for being able to digest long documents – some AI search tools (and even Bing in 2023) started leveraging Claude for certain functions (for example, Claude could potentially summarize a long webpage better). Meta’s LLaMA models, open-sourced in 2023, enabled a wave of innovation as developers built their own small-scale search/chat tools on top of them without needing access to OpenAI. We began to see community-driven Q&A bots, specialized domain chatbots (for medicine, law, etc.), all of which contribute to pulling niche queries away from general search engines. Even Elon Musk’s Grok , while initially just a chatbot on X, is a sign of more proprietary models entering the fray – xAI’s Grok was marketed as a “rebellious” chatbot with up-to-date knowledge from X, showing a different flavor that might appeal to some users over the more controlled ChatGPT ( ^[53] ) ( ^[54] ). In short, the LLM revolution lowered the barrier for entrants because models (some open, some via API) became widely available. This led to a splintering of AI sources : whereas web search used to largely mean Google’s index or Bing’s index, AI search can be powered by any number of models and data sources. For marketers and search professionals, this fragmentation means the audience’s attention is no longer monopolized by Google . One user might discover your brand via a Bing AI answer with a citation, another might hear it from a ChatGPT response (with no citation at all), and another might ask a niche AI like Character.AI or Snapchat’s My AI and get some info. It becomes crucial to monitor and optimize for multiple platforms. We will delve into strategy later in the book, but clearly one can’t assume “if we rank on Google, everyone will see us.” The rise of alternative AI search tools – even if individually small – collectively chips away at the share of queries going through any single gateway. A concrete example: Perplexity AI’s mobile app climbed app store charts in some countries in 2023, showing there is consumer demand for a “conversational search” app. And when Apple made Perplexity accessible via voice using Siri Shortcuts (basically allowing users to ask Perplexity with a voice command on iPhone), it demonstrated how easily users can pivot to a different search experience when convenience aligns. Meanwhile, the closure of Neeva indicates that not all challengers will survive – there will be consolidation or failures. But even those that fail leave an influence. Neeva’s generative answer approach, for instance, influenced Google’s and Bing’s designs. Finally, it’s worth touching on international examples of this trend. Outside the U.S., local players are also adopting LLMs in search. In China, Baidu integrated its Ernie Bot (a ChatGPT-like model) into Baidu Search in 2023, offering chat Q&A in Chinese search results. South Korea’s Naver introduced CLOVA X , an AI model for search and chat in Korean ( ^[41] ). These moves ensure that the AI revolution in search is truly global – not just Western English-centric. Each region’s dominant search engines are infusing generative AI to meet local user expectations. The effect is the same: search is becoming conversational and directly answer-oriented everywhere. In summary, the LLM-driven shake-up has invited many newcomers and approaches. While Google still stands tall, the overall search landscape is fragmenting . Users now have a menu of search options – general engines with AI (Google, Bing), AI-specific engines (Perplexity, YouChat), chatbot apps (ChatGPT, Claude, etc.), and AI assistants embedded in unrelated platforms (social media, operating systems). For the first time in decades, there is a question of “where will a user search?” every time they have a query. This pluralism challenges marketers to broaden their optimization efforts and watch emerging platforms closely. It’s a dynamic, rapidly evolving picture – and it sets the stage for the new metrics and strategies needed to track success in the era of GEO (Generative Engine Optimization), which we will explore in later chapters.

3.5 Impacts on Organic Traffic and SEO

The advent of AI-driven answers in search has had profound impacts on organic web traffic and the practice of SEO. As more queries are answered directly by large language models – either on the search results page or within chat apps – fewer users are clicking through to websites . This continuation (and acceleration) of the “zero-click search” phenomenon poses new challenges for online marketers and publishers. Let’s break down what’s happening and the data we have so far. Figure: Rising Zero-Click Searches vs. Falling Organic Traffic. A Similarweb analysis (Jul 2025) shows that after Google introduced AI Overviews (SGE) in mid-2024, the percentage of Google searches ending without any click (green line) jumped from ~56% to ~69%. In the same period, total organic search referral traffic to websites (orange line) declined significantly ( ^[55] ) ( ^[56] ). This illustrates how AI answers on the SERP are satisfying users’ queries, reducing the need to click results. Zero-click searches – where the user’s query is answered on the search page itself, so they don’t click any result – have been on the rise for years (thanks to features like featured snippets, knowledge panels, etc.). But AI answers turbocharged this trend. According to data from Similarweb, the global share of zero-click searches on Google was about 56% in early 2024 and then surged to 69% by mid-2025 ( ^[55] ) ( ^[57] ). In other words, now roughly 7 out of 10 Google searches do not result in a click to any external website. A 13-point jump in zero-click rate within a year is enormous, and this timing correlates with the rollout and increased presence of Google’s AI-generated answers (SGE). Essentially, when Google’s AI overview satisfies the query – for example, giving a multi-paragraph explanation or a list of recommendations – users feel they got what they needed and do not click additional results ( ^[58] ) ( ^[59] ). The figure above also shows the impact on traffic: total organic search traffic (visits from search engines to websites) dropped from a peak of ~2.3 billion visits per month in mid-2024 to under 1.7 billion by May 2025 ( ^[56] ). That’s a loss of over 600 million monthly visits in less than a year, presumably as users stopped clicking for answers they could read on Google’s page itself. From the perspective of publishers and SEOs, this is a worrisome trend. High-ranking content may still be used by Google’s AI to formulate an answer, but the user might never visit the site to reward it with a page view or ad impression. Google has tried to mitigate backlash by including source links in AI overviews, but they are often subtle (a small hyperlink or footnote-style number). As the Similarweb analysis put it bluntly, when the AI gives a thorough answer, “users feel they got their answer… no click needed and no traffic for the source websites.” ( ^[58] ) ( ^[59] ). This dynamic threatens traditional content models – especially for informational content where the answer can be succinctly summarized by AI. Publishers of how-to articles, travel guides, FAQ pages, and so on are seeing some traffic decline as those queries are serviced by AI summaries. It’s not just Google. Bing’s AI chat also often answers questions entirely, albeit with more prominent citations that the user can click. ChatGPT (without browsing) by design doesn’t immediately link out at all – it generates the answer from its training data. So if someone asks ChatGPT a question that your website answers, ChatGPT might just produce the answer (perhaps even paraphrasing your content if it was in the training data) and the user never visits your site. A Semrush study found ChatGPT had sent referral traffic to about 30,000 unique domains by late 2024 via its browsing tool ( ^[60] ), which shows there is some traffic coming from AI, but that number is modest relative to the scale of search traffic Google sends. It also found that ChatGPT directly answered 54% of queries without needing a search at all ( ^[61] ). So more than half the time, ChatGPT satisfied the user entirely internally. The remaining cases where ChatGPT did a search or the user clicked a source link account for far fewer total clicks than an equivalent volume of Google searches would generate. The net effect is clear: the rise of AI Q&A has eaten into the clicks that websites used to get from queries . That said, not all is doom and gloom. In some AI search experiences, there are still opportunities to capture traffic. Bing Chat and Perplexity both include citations as a core part of their design, encouraging users to click through for more details. Many users of Bing’s chat do click the footnote links to verify information or read further (though Microsoft hasn’t shared exact percentages publicly). Perplexity AI often presents its answers with multiple source links in line, acting almost like a portal – a user might read the summary then click two of the cited articles for deeper reading. In that sense, certain AI-driven platforms can become new referral sources . For example, if your site is frequently cited by Perplexity for queries in your niche, you might see traffic coming from perplexity.ai URLs. Some publishers have reported small but noticeable traffic from Bing’s new Bing (chat) as well, since Bing often attributes content. The Semrush analysis highlighted that tech and education sites in particular got increased referral traffic from ChatGPT’s browsing mode in late 2024 ( ^[62] ) – likely because a lot of ChatGPT’s users were asking coding and academic questions that caused the bot to fetch content from documentation sites, Stack Overflow, arXiv papers, etc., and sometimes users clicked through. Furthermore, Google’s SGE citations – while perhaps reducing clicks – still drive some traffic for in-depth content. Google often presents just a snippet in the AI overview and then shows a “ More about this ” link or visual element that, when clicked, leads to the source. Engaged users who want the full context may still click those links. Google has also indicated that if its AI overview is drawn from multiple sources, it tries to give each due credit, which could distribute some clicks. However, one worrying sign: early observations showed that when the AI overview is very comprehensive, users skip even the top organic results that appear below, leading to across-the-board decline in click-through rate (CTR) . A study in late 2024 found that click-through rates on Google ads and organic results dropped notably for longer, detailed queries where AI snapshots were likely to appear ( ^[63] ) ( ^[64] ). This aligns with the idea that the more detail in the query (and thus the more detailed the AI answer), the fewer clicks needed because the answer is right there. The overall impact on SEO strategy is significant. Traditional SEO has long chased positions that get clicks (position #1, featured snippets, etc.). Now, one must consider AI visibility : is your content being referenced or used by the AI answers? If so, you may still reach the audience (in a brand awareness sense) even if you don’t get the click. For example, if an AI summary says “According to YourSite ,… [some info]” (as Perplexity or Bing might output explicitly), the user sees your brand name. This has value in itself, potentially driving branded searches or direct visits later, even if that particular session was zero-click. In Google’s case, it typically doesn’t mention brand names in the overview text (it just footnotes), but savvy users might notice the source. There’s also the aspect of trust : if users become wary of AI accuracy, they might click sources more often to verify. Some fraction of users do this, especially for important decisions – they treat the AI as a starting point and then visit one of the cited sites to double-check. SEO practitioners can cater to that by ensuring their brand/site is among those cited, increasing the odds of getting that verifying click. Another repercussion is on the type of content that gets traffic. If an AI can answer a simple factual question (e.g. “What’s the capital of Botswana?”), the website that used to get that traffic (like a Wikipedia page or travel site) will now see far fewer hits for that query. On the other hand, more complex, subjective, or up-to-the-minute queries might still drive clicks. AI sometimes struggles or avoids giving answers on very recent news (due to limited training cutoff or guardrails), instead pointing users to news sources. Google’s SGE, for instance, usually does not generate an AI answer for breaking news or specific site navigational queries – it defaults to normal results. So news publishers aren’t directly seeing AI-overview cannibalization (Google even confirmed SGE is limited for news), but they are seeing an indirect decline if overall search usage shifts or if fewer people search news because they got some summary in social media or via an AI elsewhere. In fact, Similarweb data shows news publishers’ organic traffic fell sharply in the period after SGE launch ( ^[65] ) ( ^[56] ). Likely reasons: fewer searches about topics that AI can handle, or users asking AI directly for summaries of news. For example, someone could ask Bard “What’s the latest on the company X acquisition rumor?” and Bard might compile info from multiple news articles, satisfying the curiosity without the user visiting each news site. Marketers are adapting in several ways. Firstly, optimizing for featured snippets and structured data has taken on new meaning – it’s not just about getting the snippet but also about being the content the AI pulls. Ensuring content is well-structured, factual, and authoritative increases the chances that the AI will use it (Google’s systems and Bing’s algorithms still rely on high-quality sources to feed the AI). There’s a renewed emphasis on E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) because the AI engines want trustworthy info to avoid mistakes. If your content is seen as authoritative (good backlinks, user engagement, schema markup for clarity, etc.), it’s more likely to be chosen by the generative search algorithms to include in answers ( ^[66] ) ( ^[67] ). Some companies are even explicitly weaving likely Q&A phrasing into their content (anticipating how an AI might need to quote a self-contained answer). Secondly, measurement metrics are shifting . SEO success can’t be measured only by clicks and traffic in the AI age. Marketers are starting to look at brand mentions in AI (even unlinked) and overall visibility. If an AI answer consistently lists your product as “one of the top options” for certain queries (say, “best CRM software” queries answered by Bing AI include your brand via a citation from a review site), that has marketing value akin to a top ranking – even if the user doesn’t click immediately, they’ve seen your brand. We’ll explore this more in Chapter 13 on measurement, but it’s worth noting now that qualitative impact (influence, awareness) from AI answers is a new frontier. Some tools are emerging to track if and how often a brand or URL gets mentioned in various AI responses ( ^[68] ). Another impact is that content strategy might pivot toward topics and formats that AI can’t easily cannibalize . For instance, AI has trouble with very personal experiences, original research, or real-time community discussions. So some publishers focus on content with a strong personal voice or proprietary data – content that an AI overview would either skip or have to direct the user to. Also, interactive tools and features on websites (calculators, quizzes, etc.) remain unreplicated by static AI answers and can draw clicks. Essentially, sites may differentiate to give users reasons to still visit. Finally, the industry is grappling with the balance between providing data to AI vs. holding it back. Some publishers, notably those in forums like Stack Overflow or some news outlets, felt so much traffic drop that they considered blocking AI crawlers to prevent free use of their content. In mid-2023, Stack Overflow saw a nearly 50% drop in traffic compared to 2022 ( ^[69] ), which coincided with developers using ChatGPT for coding questions (and Stack Overflow also had content moderation issues). While the decline wasn’t solely due to AI (Stack Overflow had other community issues ( ^[70] )), AI was “throwing fuel on the fire” as the Medium analysis said ( ^[70] ). This kind of drastic decline raises questions: should these sites allow AI models to scrape their content to give away answers? Some moved to restrict OpenAI’s GPTBot in robots.txt, or explore paid partnerships (e.g., Reddit and Stack Exchange signaled they want compensation for AI training on their data). This is an evolving battleground that blurs the line between SEO and broader data policy. As of 2024, no clear solution has emerged, but it’s part of the new SEO reality that your content could be consumed by millions via AI with little credit or traffic – and strategies must account for that. In conclusion, the LLM revolution has intensified the long-brewing challenges of zero-click search and altered how success is defined in organic visibility. Marketers now must optimize not just for the blue links , but for the answer boxes – both the ones users see and the ones AI models “see” when training or retrieving information. While overall organic traffic may be trending downward due to AI answers, those who adapt by ensuring their content is AI-friendly (in terms of being used/cited) and by focusing on areas where human touch is irreplaceable can still thrive. The next chapters will dive deeper into strategies (technical and content-wise) for doing exactly that – but first, having grounded ourselves in how search behavior and platforms have transformed by 2025, we’ve set the stage for why GEO (Generative Engine Optimization) is now a critical discipline alongside traditional SEO. The “LLM revolution” in search is here, and its effects on traffic and tactics are impossible to ignore.

References

^[1] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[2] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[3] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[4] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[5] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[6] www.aiprm.com - Aiprm.Com URL: https://www.aiprm.com/chatgpt-statistics

^[7] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/keyword-query-length-insights-445376

^[8] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/keyword-query-length-insights-445376

^[9] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[10] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[11] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[12] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[13] Explodingtopics.Com Article - Explodingtopics.Com URL: https://explodingtopics.com/blog/chatgpt-users

^[14] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[15] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[16] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[17] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[18] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/GPT-4

^[19] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/GPT-4

^[20] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/GPT-4

^[21] www.mindandmetrics.com - Mindandmetrics.Com URL: https://www.mindandmetrics.com/blog/ai-retrospective-ai-trends-from-2023-and-predictions-for-2024

^[22] www.mindandmetrics.com - Mindandmetrics.Com URL: https://www.mindandmetrics.com/blog/ai-retrospective-ai-trends-from-2023-and-predictions-for-2024

^[23] www.impressiondigital.com - Impressiondigital.Com URL: https://www.impressiondigital.com/blog/bing-differ-google

^[24] www.forbes.com - Forbes.Com URL: https://www.forbes.com/sites/johnkoetsier/2023/02/17/since-adding-chatgpt-bings-app-grew-almost-as-much-as-all-last-year

^[25] Timesofindia.Indiatimes.Com Article - Timesofindia.Indiatimes.Com URL: https://timesofindia.indiatimes.com/gadgets-news/chatgpt-hasnt-helped-microsoft-bing-find-the-right-answers-to-beat-google-search/articleshow/106981810.cms

^[26] Timesofindia.Indiatimes.Com Article - Timesofindia.Indiatimes.Com URL: https://timesofindia.indiatimes.com/gadgets-news/chatgpt-hasnt-helped-microsoft-bing-find-the-right-answers-to-beat-google-search/articleshow/106981810.cms

^[27] Timesofindia.Indiatimes.Com Article - Timesofindia.Indiatimes.Com URL: https://timesofindia.indiatimes.com/gadgets-news/chatgpt-hasnt-helped-microsoft-bing-find-the-right-answers-to-beat-google-search/articleshow/106981810.cms

^[28] Timesofindia.Indiatimes.Com Article - Timesofindia.Indiatimes.Com URL: https://timesofindia.indiatimes.com/gadgets-news/chatgpt-hasnt-helped-microsoft-bing-find-the-right-answers-to-beat-google-search/articleshow/106981810.cms

^[29] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/GPT-4

^[30] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/GPT-4

^[31] www.mindandmetrics.com - Mindandmetrics.Com URL: https://www.mindandmetrics.com/blog/ai-retrospective-ai-trends-from-2023-and-predictions-for-2024

^[32] www.mindandmetrics.com - Mindandmetrics.Com URL: https://www.mindandmetrics.com/blog/ai-retrospective-ai-trends-from-2023-and-predictions-for-2024

^[33] www.adweek.com - Adweek.Com URL: https://www.adweek.com/media/google-ai-search-ad-business-openai-perplexity

^[34] www.adweek.com - Adweek.Com URL: https://www.adweek.com/media/google-ai-search-ad-business-openai-perplexity

^[35] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[36] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[37] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[38] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[39] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[40] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[41] Pmc.Ncbi.Nlm.Nih.Gov Article - Pmc.Ncbi.Nlm.Nih.Gov URL: https://pmc.ncbi.nlm.nih.gov/articles/PMC11091448

^[42] www.adweek.com - Adweek.Com URL: https://www.adweek.com/media/google-ai-search-ad-business-openai-perplexity

^[43] www.macrumors.com - Macrumors.Com URL: https://www.macrumors.com/2025/05/07/apple-working-on-ai-search-in-safari

^[44] www.macrumors.com - Macrumors.Com URL: https://www.macrumors.com/2025/05/07/apple-working-on-ai-search-in-safari

^[45] www.macrumors.com - Macrumors.Com URL: https://www.macrumors.com/2025/05/07/apple-working-on-ai-search-in-safari

^[46] www.macrumors.com - Macrumors.Com URL: https://www.macrumors.com/2025/05/07/apple-working-on-ai-search-in-safari

^[47] www.jasper.ai - Jasper.Ai URL: https://www.jasper.ai/blog/what-is-perplexity-ai

^[48] www.jasper.ai - Jasper.Ai URL: https://www.jasper.ai/blog/what-is-perplexity-ai

^[49] www.jasper.ai - Jasper.Ai URL: https://www.jasper.ai/blog/what-is-perplexity-ai

^[50] www.jasper.ai - Jasper.Ai URL: https://www.jasper.ai/blog/what-is-perplexity-ai

^[51] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[52] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Grok_ (chatbot

^[53] www.cbsnews.com - Cbsnews.Com URL: https://www.cbsnews.com/news/elon-musk-grok-4-ai-chatbot-x

^[54] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2025/7/10/what-is-grok-and-why-has-elon-musks-chatbot-been-accused-of-anti-semitism

^[55] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[56] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[57] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[58] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[59] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[60] www.digitalinformationworld.com - Digitalinformationworld.Com URL: https://www.digitalinformationworld.com/2025/02/chatgpt-answers-54-of-queries-without.html

^[61] www.digitalinformationworld.com - Digitalinformationworld.Com URL: https://www.digitalinformationworld.com/2025/02/chatgpt-answers-54-of-queries-without.html

^[62] www.digitalinformationworld.com - Digitalinformationworld.Com URL: https://www.digitalinformationworld.com/2025/02/chatgpt-answers-54-of-queries-without.html

^[63] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/keyword-query-length-insights-445376

^[64] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/keyword-query-length-insights-445376

^[65] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/similarweb-zero-click-search-surge-google-ai-overviews-3562

^[66] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[67] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/chatgpt-search-insights

^[68] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/keyword-query-length-insights-445376

^[69] Medium.Com Article - Medium.Com URL: https://medium.com /@ThreadSafeDiaries/stack-overflows-traffic-dropped-50-here-s-the-real-reason-6efc12c312c4

^[70] Medium.Com Article - Medium.Com URL: https://medium.com /@ThreadSafeDiaries/stack-overflows-traffic-dropped-50-here-s-the-real-reason-6efc12c312c4

4. How Large Language Models Work (What Marketers Should Know)

Large Language Models (LLMs) like OpenAI’s GPT-4, Google’s Gemini, Anthropic’s Claude, Meta’s Llama 2, and the new xAI Grok are transforming how information is found and delivered. Unlike traditional search engines that index and retrieve web pages, LLM-powered tools generate answers by predicting text based on patterns learned from vast training data. This fundamental difference has big implications for digital marketing and SEO strategy. In this chapter, we’ll explore how LLMs work – from their training and knowledge limitations to issues like hallucination, memory, and relevance – and what marketers need to know to optimize content in this new era of Generative Engine Optimization (GEO).

4.1 Training Data and Language Prediction

Modern LLMs are built on transformer neural network architectures and are trained on enormous datasets of text (web pages, books, articles, forums, and more). Models such as GPT-4, Google’s Gemini, and Meta’s Llama owe their capabilities to ingesting hundreds of billions of words from diverse sources ( ^[1] ). During training, the model learns to predict the next word in a sentence, over and over, across countless examples. In essence, an LLM develops a statistical understanding of language: it doesn’t search for answers in a database at query time, but rather synthesizes a likely answer word-by-word based on the patterns it absorbed during training. For marketers, this distinction is crucial. It means that unlike a search engine that might surface your exact webpage if it’s deemed relevant, an LLM might generate an answer using information from your content (perhaps paraphrased or summarized) without directly quoting or linking to it. The model focuses on producing a fluent, relevant response – not on attributing sources by default. Training on vast text corpora – OpenAI’s GPT-4, for example, was pre-trained on “publicly available internet data” and additional licensed datasets, with content spanning everything from correct and incorrect solutions to math problems to a wide variety of ideologies and writing styles. This broad training helps the model answer questions on almost any topic. Google’s Gemini, similarly, has been described as a multimodal, highly general model drawing on extensive text (and even images and code) in its training. Meta’s Llama 2 was trained on data like Common Crawl (a massive open web archive), Wikipedia, and public domain books. In practical terms, these models have read a large portion of the internet. They don’t have a simple index of facts, but rather a complex probabilistic model of language. One implication is that exact keywords matter less in LLM responses than overall content quality and clarity. Since an LLM generates text by “predicting” a reasonable answer, it might not use the exact phrasing from any single source. This means your content could influence an AI-generated answer even if you’re not explicitly quoted, provided your material was part of the model’s training data or available to its retrieval mechanisms. It also means an LLM can produce answers that blend knowledge from multiple sources. For example, ChatGPT might answer a question about a product by combining a definition from one site, a user review from another, and its own phrasing – all without explicitly telling the user where each piece came from. As a marketer, you can’t assume that just because you rank #1 for a keyword, an LLM will present your content verbatim to users. Instead, the model might absorb your insights into a broader answer. This elevates the importance of writing content that is clear, factually correct, and semantically rich, because LLMs “care” about coherence and usefulness more than specific keyword frequency. LLMs synthesize rather than retrieve. Traditional search engines retrieve exact documents and then rank them. LLMs like GPT-4 generate a new answer on the fly. They use their training to predict what a helpful answer sounds like . As an analogy, think of an LLM as a knowledgeable editor or author drafting a new article based on everything they’ve read, rather than a librarian handing you an existing book. This is why LLM answers can sometimes feel more direct or conversational – the model is essentially writing the answer for you. It’s also why errors (hallucinations) can creep in, which we’ll discuss later. From an SEO perspective, this generative approach means that high-quality, well-explained content stands a better chance of being reflected in AI-generated answers than thin content geared solely to rank on Google. The model might not use your exact words, but if your page provides a clear, authoritative explanation, the essence of that information may inform the AI’s response. Conversely, stuffing pages with repetitive keywords or SEO gimmicks is less effective, because the LLM isn’t indexing pages by keyword; it’s absorbing content for meaning and then later recalling the meaning more than the literal words. It’s worth noting that LLMs are extremely large – GPT-4 is estimated to have over 1.7 trillion parameters (the internal weights that store learned patterns). These parameters encode probabilities for word sequences. When an LLM answers a query, it starts with the user’s prompt, and then it internally predicts a sequence of words that statistically and contextually fit. The self-attention mechanisms in transformers allow the model to consider relationships between words and concepts even if they are far apart in the text. For example, if a user asks “How does a hybrid car work?”, the model doesn’t search for a specific document. Instead, it uses its trained neural connections (built from seeing millions of words about cars) to produce an explanation, perhaps describing the battery, electric motor, and gasoline engine in seamless prose. It might have “seen” text about hybrid cars during training, but it’s not copying one source – it’s generating new sentences that sound like what it learned. This ability to synthesize means content creators should focus on providing comprehensive coverage of topics in a way an AI can easily learn from. In other words, ensure your content teaches well – because the better the AI can learn your information, the more likely it is to use it when generating answers. Why this matters for marketers: In the old SEO paradigm, one might obsess over exact-match keywords or getting a featured snippet. In the new GEO paradigm (Generative Engine Optimization), the emphasis shifts to being part of the model’s knowledge base and providing information in a format the model can easily understand and repurpose. If your content is high-quality and clearly written, an LLM is more likely to have that knowledge and use it. If your content is thin, misleading, or overly optimized just for search crawlers, an LLM may either ignore it or, worse, learn incorrect information from it (which could later reflect poorly in AI outputs). The bottom line is that LLMs reward clarity and depth. As one SEO expert put it, structured writing and genuine semantic clarity are not optional in the age of generative AI – they are essential. LLMs don’t look for a <meta> tag to figure out your page’s topic; they literally read your content like a very fast, very well-read user. Thus, good writing and organization become your new SEO superpowers.

4.2 Knowledge Cutoffs vs. Real-Time Data: Retrieval-Augmented Generation (RAG)

One limitation of many LLMs is the knowledge cutoff – the point in time after which the model has seen no new training data. For instance, the base GPT-4 model (as of early 2024) has a knowledge cutoff around September 2021. This means if you ask it about an event or statistic from 2022 or 2023, it might not know about it from its training alone. Similarly, Meta’s Llama 2 has a training cutoff of late 2022. This poses a challenge: users expect up-to-date information, but these models’ “memory” can be frozen in time. To address this, AI developers have introduced real-time retrieval mechanisms that supplement the static knowledge of LLMs with fresh information. This approach is broadly known as Retrieval-Augmented Generation (RAG) . In a RAG system, when the user asks a question, the AI first performs a search or lookup in an external data source (for example, a web search index, a company knowledge base, or a database of documents) and retrieves relevant text. That retrieved text is then fed into the LLM along with the user’s query, giving the model “grounding” facts to base its answer on ( ^[2] ). The LLM then generates a response that incorporates that up-to-date information. Essentially, RAG combines the strengths of search (accurate, current data retrieval) with the strengths of LLMs (fluent natural language answers). Examples of RAG in action: If you use Bing Chat , you’re seeing RAG at work – Bing’s AI (which uses GPT-4 under the hood) will actually search the live web for your query, then use the web results to formulate an answer, often citing sources. Another example is Perplexity.ai , an AI search engine that always provides citations: when you ask Perplexity a question, it finds relevant up-to-date sources (e.g. news articles, websites) and then generates a concise answer with footnotes linking to those sources. Google’s Search Generative Experience (SGE) , currently in preview, also uses live search results to generate an “AI overview” at the top of the page for certain queries. This overview is built by the LLM reading top search hits and synthesizing them. In all these cases, the AI is not limited to its stale training data – it can pull in new information on the fly. An example of real-time AI retrieval: Perplexity.ai answers a query about the latest Madonna concert setlist by fetching current info (March 2024) and citing trusted sources. This Retrieval-Augmented Generation approach ensures up-to-date, factual answers, addressing the knowledge cutoff problem. For marketers, RAG is a double-edged sword. On one hand, it means fresh content can be surfaced by AI even if that content wasn’t in the model’s original training. If you publish a blog post tomorrow and it starts ranking or is deemed relevant, an AI like Bing or Perplexity might pull it in to answer user questions next week. This is encouraging – it preserves some role for traditional SEO (you still want to appear in those top results that the AI will consider). On the other hand, if AI platforms are giving users the answers directly, the click-through to your site may be reduced . We’ll discuss metrics in a later chapter, but it’s important to recognize that RAG-driven answers often satisfy the user without a click (especially if the answer is fully contained with citations). Your content’s value might be realized by informing the AI’s answer rather than driving traffic. This makes brand visibility (being mentioned or cited by the AI) a new key goal alongside traditional clicks. Marketers should also understand how RAG selects information . Typically, a retrieval algorithm (which might be a traditional keyword-based search or a vector similarity search) finds text passages that likely answer the user’s query. Those passages are then given to the LLM. Importantly, even sophisticated AI search still often relies on keyword matching for the retrieval step. In other words, to be one of the sources an AI pulls in, your content likely needs to contain the keywords or phrases the user’s query uses (or very close synonyms). The LLM itself can understand nuanced content, but if the retrieval mechanism doesn’t surface your page, the model won’t even see your content. As one analysis noted, the “retrieval layer” that decides what content is eligible to be summarized is still driven by surface-level language cues – in experiments, simple keyword-based retrieval (BM25) outperformed purely semantic approaches for feeding documents to the LLM. In plain terms: even for AI-generated answers, classic keyword strategy isn’t dead. Clear, literal wording that matches user queries helps ensure your content is selected as input to the AI. So while LLMs themselves don’t need exact-match keywords to understand text, the pipeline bringing them content does often depend on those keywords ( ^[3] ). A user prompt like “Show me articles about LLMs using schema” will cause the system to fetch content that explicitly mentions “LLMs” and “schema,” not just content that implies those concepts ( ^[3] ). This means you should still align your content with the language your audience uses in queries. Another aspect of staying visible in the age of RAG is ensuring your content is accessible to AI crawlers and indexes . For example, OpenAI introduced GPTBot , a web crawler that browses the internet to collect content for future model training. GPTBot honors robots.txt ; by default it will crawl sites to gather data that could be used in the next GPT model. Some website owners have chosen to block GPTBot due to privacy or IP concerns. As of mid-2025, over 3% of websites globally (and a larger share of top sites) disallow GPTBot. This is an important strategic decision. If you block AI crawlers from training on your content, your information might not be present in the next generation of models. That could limit your brand’s visibility in AI answers. It’s a trade-off: some publishers worry about giving content away to AI without direct compensation or traffic, while others see being included in AI training sets as a way to ensure their brand knowledge is widespread. As noted in an industry discussion, blocking GPTBot “restricts your content from being used in AI-generated responses, which can limit brand visibility in tools that now dominate early-stage discovery.” Conversely, allowing it means “your brand [can] show up in ChatGPT answers,” potentially reaching a massive user base. In fact, ChatGPT reportedly reached around 800 million weekly users at one point – a staggering audience you’d probably want your content to be exposed to. Making content accessible goes beyond GPTBot. It also means continuing to allow indexing by search engines (since AI search experiences like SGE or Bing are built on top of search indexes) and ensuring your content isn’t buried behind logins or paywalls (unless your business model demands it). If you have a developer API or data feed, you might even consider making certain data available to trusted AI partners – for instance, some e-commerce sites might feed product info to Bing’s index so that Bing Chat can answer product questions with live data. The key point is that content availability equals AI visibility . Marketers should keep an eye on emerging standards for AI inclusion/exclusion (similar to robots.txt but for AI). For now, the pragmatic approach to GEO is to welcome reputable AI crawlers : this can help AI models stay up-to-date on your offerings and reduce the risk of misinformation (since the AI will have your latest, correct info to learn from). It’s also worth mentioning that some AI platforms (OpenAI, Meta, etc.) do periodically retrain or fine-tune their models with newer data. OpenAI has hinted at GPT-4 updates that include some post-2021 knowledge via fine-tuning or the plugin/browsing features. Google’s Gemini is likely trained on more recent data (possibly through 2023) and Google can continuously infuse freshness via its search index. Anthropic’s Claude regularly gets fine-tuned with more recent content as well. In the enterprise space, companies are deploying internal RAG systems that connect LLMs to their up-to-the-minute databases. All this means the gap between what happened today and what the AI knows is closing , but not entirely gone. Marketers should still produce timely content (AI will find ways to use it via retrieval) and also evergreen content (to be included in base training sets and long-term AI knowledge). Ensure that when the next wave of model training happens, your site is crawlable and your content is high-quality – that maximizes the chance that the AI will “learn” your content. OpenAI’s own GPT-4 documentation notes that its training data included a diverse mix intended to capture “recent events” and “strong and weak reasoning” etc., but it admitted a knowledge cutoff in 2021. With GPT-5 or others, those cutoffs will extend. The concept of GEO includes being prepared for your content to be training data. For instance, if your site offers an FAQ or glossary in your niche, that’s exactly the kind of text likely to be scooped up and learned by LLMs (because it’s explanatory and authoritative). In a sense, content SEO and “training SEO” become one : you write for users and for the AIs that read over the users’ shoulders. In summary, RAG and real-time data integration are bridging the gap between static AI knowledge and the current world . Marketers must adapt by (a) keeping content accessible and indexable to these systems, (b) continuing to optimize for relevant keywords and clear language so that retrieval algorithms can find you, and (c) recognizing that being the source of truth in AI answers (even if indirectly) is the new win. A practical tip: monitor where your content might be appearing in AI citations. For example, if Perplexity or Bing Chat often cites your blog, that’s a good sign your GEO strategy is working. Some SEO tools now even track “AI mentions” or how often an AI assistant references a brand. We’ll cover measurement in Chapter 13. Before moving on, it’s important to note that RAG isn’t just about freshness – it’s also a solution for accuracy , which leads us into the next topic. By grounding answers in real sources, RAG significantly reduces the incidence of AI hallucination and increases trust (users can see citations). Let’s discuss hallucinations and why factual accuracy is a critical concern.

4.3 Hallucinations and Accuracy Challenges

One of the most notorious quirks of LLMs is their tendency to “hallucinate” – in other words, to produce information that sounds confident and specific, but is completely made up or incorrect. Unlike a search engine that simply might not return a result if it doesn’t have one, an LLM will always try to answer your question. If the model doesn’t actually know something, it will improvise, drawing on its training patterns to create a plausible-sounding answer. This can range from minor inaccuracies (getting a date or name wrong) to completely fabricated facts, citations, or even quotes. For individuals and brands, hallucinations aren’t just academic errors – they can be reputational or legal landmines. Consider a real-world example: in April 2023, an Australian mayor discovered that ChatGPT was mistakenly stating he had been involved in a bribery scandal and even served prison time – none of which was true. The mayor was actually the whistleblower who exposed that scandal, not a perpetrator. He was understandably alarmed and pursued what could be the first defamation lawsuit against OpenAI for this false claim. ChatGPT had essentially hallucinated a criminal history for a real person, potentially damaging his reputation. The mayor’s lawyers noted how the AI’s answer gave a false sense of authority – because ChatGPT does not cite sources by default, an average user might just assume the information is correct. This case illustrates the brand risk: an AI could incorrectly describe your company or a public figure associated with you, and users might believe it because of the AI’s confident tone. Another well-known incident: In mid-2023, a pair of New York lawyers filed a legal brief that cited six precedent court cases – all of which were fake, invented by ChatGPT. The lawyers had used ChatGPT to research cases, and the AI provided entirely fictitious case names and summaries that sounded authentic (complete with docket numbers and judges’ names). The judge was not amused; he sanctioned the attorneys with a fine and reprimand. The lawyers admitted they never imagined the AI would just make up cases “out of whole cloth”. This example underscores that even highly educated professionals can be misled by AI hallucinations if they aren’t careful. It also highlights a crucial point: LLMs do not have an internal database of verified facts – their knowledge is statistical, not deterministic. Without external verification, they might output incorrect information that looks perfectly credible. From a marketing perspective, hallucinations pose a risk to brands and a challenge to AI adoption . If an AI chatbot erroneously states something about your product (say, it hallucinates a feature that doesn’t exist or confuses your product with a competitor’s), customers could be misinformed. Or the AI might misquote a statistic from your content, undermining your thought leadership with inaccuracies. Even more concerning, if the AI has absorbed biased or untrue statements about your brand from somewhere, it might repeat them. Brands have already started monitoring AI outputs for such mentions. For example, if you’re a PR manager, you might now need to check not just Google results for your brand, but also ask ChatGPT or Bing Chat “What does [Brand] do and is it reliable?” to see if the AI says anything incorrect or damaging. Why do hallucinations happen? At a technical level, it’s because LLMs are optimized to produce fluent language that seems right, rather than to internally fact-check against a knowledge graph. The model is driven by probabilities – it knows what words often co-occur. If asked a question it’s unsure about, it will generate something that usually would be an answer. It has no intrinsic concept of truth, only what it learned during training. If the training data was sparse or conflicting on that point, or if the query is very specific, the model may just pick a likely-sounding completion. For instance, early versions of Google’s Bard made a notable mistake when asked about the James Webb Space Telescope – Bard confidently gave an incorrect fact, which in a high-profile demo led to criticism and even a dip in Google’s stock price. The model wasn’t trying to lie; it just didn’t know the correct answer and guessed. Some metrics show the scale of the issue: in a study examining LLM responses in a scientific context, GPT-3.5 was found to produce a hallucinated reference (a made-up citation) about 40% of the time, and even GPT-4 did so about 28% of the time. Google’s Bard, in that 2023 study, hallucinated references a whopping 91% of the time. While those numbers may vary by context and have likely improved with model updates (and Bard has since been upgraded, possibly via Gemini), the takeaway is that even the best models currently in use are far from perfectly accurate . They do make things up. GPT-4 is more reliable than its predecessors (OpenAI claims it reduces hallucinations significantly, and indeed GPT-4’s hallucination rate is lower than GPT-3.5’s), and newer iterations (like a hypothetical GPT-4.5 or GPT-5) are expected to further improve. Anthropic’s Claude has been designed with constitutional AI principles to avoid incorrect statements, and users often report it has a slightly different style that can reduce certain errors. But no LLM is 100% factual. Marketers must therefore approach AI content generation with a critical eye: AI can accelerate content creation and user interaction, but its outputs must be verified, especially on factual details. What can be done about hallucinations? There are a few approaches, and many tie back to content strategy: Authoritative, well-structured content: If your website clearly and unambiguously states facts about your domain, an LLM is less likely to hallucinate when using your content. Conversely, if correct information is scarce or drowned in a sea of speculation online, the model may latch onto the wrong patterns. This is why one recommendation is to publish fact sheets, Q&As, and data pages about your brand or industry. By seeding the web (and thus training data) with accurate information, you help steer the AI’s model. It’s akin to SEO in the sense of providing the canonical answer for your area of expertise. Retrieval and citations (RAG): As covered above, retrieval-augmented generation can drastically cut down hallucinations. When the model is forced to consult external sources (like a live database or a snippet from your site) before answering, it is more likely to stay factual ( ^[2] ). That’s why tools like Bing Chat or Perplexity that cite sources tend to inspire more confidence – if the AI says “According to [Source], the product weighs 1.2 kg,” and gives you the source, you trust it more and the chance of a total fabrication is lower. Marketers integrating AI into user experiences (e.g., a website chatbot) should strongly consider a RAG approach: have the bot pull answers from your knowledge base or site content, rather than relying purely on its pre-trained memory. Not only does this improve accuracy, it also allows you to update information immediately (update the knowledge base and your AI assistant will reflect that, without needing a full model retrain). Human oversight and fact-checking: When using generative AI for content creation, implement a review process. If AI writes a draft of an article, have a subject matter expert review every claim and statistic. AI can save you time by generating well-structured text, but you must ensure it hasn’t introduced a false “fact.” For instance, if you prompt ChatGPT to write “10 Benefits of Product X,” it might invent a benefit that isn’t real if it runs out of sourced material. It’s up to you to catch that. Treat AI outputs as you would a human junior copywriter’s work – useful, but requiring editorial oversight. Model improvements: The AI research community is aware of hallucination issues and is actively working on them. Techniques like reinforcement learning from human feedback (RLHF) have been used to fine-tune models to be more truthful. OpenAI, for example, had humans vote on preferred answers which presumably included preferring correct over incorrect answers, thereby nudging GPT-4 to lie less. Other approaches involve adding modules for verification – e.g., after the model generates an answer, have it check the answer against a trusted source (sort of a self-RAG). While these innovations are promising, from a marketer’s standpoint it’s safer to assume the AI will make mistakes and plan accordingly, rather than waiting for a “perfectly honest” model. The implications for brands are also prompting conversations about governance and liability . If an AI platform repeatedly hallucinates harmful falsehoods about businesses or people, will there be legal repercussions? The Australian mayor’s threatened lawsuit is a test case. In another case in the US, a radio host sued OpenAI after ChatGPT falsely accused him of embezzling money – that case was dismissed on the grounds that OpenAI itself didn’t publish the info (someone’s usage of ChatGPT did), but we’re in uncharted territory legally. Marketers should thus monitor AI outputs for their brands similarly to how they monitor social media or press mentions. We might see the rise of “AI Reputation Management” as a field. On the flip side, there are opportunities: ensuring your brand has a strong, positive presence in the data that AIs train on (through content marketing, PR, etc.) could help the AI tell your story correctly. For example, if you publish an open dataset or detailed history about your company, a future LLM might learn from it and convey that information to users accurately, rather than pulling from a dubious blog post written by someone else. In summary, hallucinations are a current reality of LLMs – they can and do fabricate information . This elevates the importance of authoritative content and fact-checking. Brands should double down on being the source of truth in their domain. By doing so, you reduce the chance that an AI fills a knowledge gap with nonsense. And when using AI outputs, maintain a healthy skepticism. As one AWS expert quipped, an LLM on its own can be like “an over-enthusiastic new employee who refuses to stay informed with current events but will always answer every question with absolute confidence.” You wouldn’t let a new hire present to clients unsupervised on day one – likewise, let’s not let AI outputs go live without a sanity check. The goal is to harness LLMs’ productivity and conversational power while safeguarding accuracy – a balance that is still being learned industry-wide.

4.4 Context and Memory: Multi-Turn Conversations

A major evolution from traditional search to LLM-based chat interfaces is the concept of contextual, multi-turn conversation. In a normal search engine, each query you type is independent – the search engine doesn’t remember what you asked 5 minutes ago. In contrast, when you interact with an AI chatbot (be it ChatGPT, Bing Chat, Google’s Bard/Gemini, or others), the system retains memory of the dialogue (up to certain limits) and uses that context to inform subsequent responses. This fundamentally changes how users seek information and how content might be consumed or referenced over multiple turns. LLMs remember the conversation (up to a point). Technically, this is handled via the “context window” of the model – a rolling buffer of the last N tokens of dialogue that the model takes into account. Early versions like GPT-3 had context windows of around 4,000 tokens (roughly 3,000 words). Newer models have expanded this greatly: OpenAI’s GPT-4 offers variants with up to 32,000 tokens (24k words) of memory, and Anthropic’s Claude went even further with a staggering 100,000-token context window (approximately 75,000 words). To put that in perspective, Claude can ingest an entire novel or a lengthy research report in one go and discuss it. Claude’s creators demonstrated this by feeding it the full text of The Great Gatsby (72K tokens) and asking a detailed question about a subtle change – Claude answered correctly in seconds. This long memory enables conversations that can reference a large document or many prior messages without losing track. For marketers, the immediate implication is that AI chatbots can handle complex, in-depth queries that build on each other.

A user might start general (“What are the benefits of electric cars?”), then follow up with something more specific (“How does the maintenance cost compare to hybrid cars?”), and then maybe, “You mentioned battery lifespan, can you provide more details on that for a Nissan Leaf?” In a chat scenario, the AI will carry forward all relevant information from earlier in the chat when answering the later questions. It behaves more like a human advisor who remembers what you already asked and what they already told you. So how do we optimize content knowing that user interactions might be multi-turn and contextually layered? Here are a few considerations: Chunk your content into logical sections that can stand alone. Since the AI may not present an entire webpage to the user but rather use pieces of it across different turns, it helps if each section of your content addresses a specific subtopic clearly. For instance, if you have a product FAQ page, ensure each Q&A pair is self-contained. If the chatbot draws on a particular Q&A to answer one question, and then the user asks a follow-up, the AI might go back to the same source or related ones. If your content is written in long, interwoven paragraphs that mix many ideas, the AI might extract an incomplete snippet that doesn’t carry the full context to the next turn. Conversely, if each paragraph or section covers one idea succinctly, the AI can quote or summarize that section when needed without misrepresenting it. Use consistent terminology and references. In a conversation, pronouns and references matter.

For example, if your content refers to “Product X” in one paragraph and then “the device” in the next, a human understands those are the same, but an LLM might need clear signals to maintain reference across turns. In a multi-turn dialogue, the AI uses its memory of previous mentions. If a user asks about “the device” later, the AI will try to link that to “Product X” mentioned earlier, but clarity helps. Make sure your content uses names and terms clearly so that if an AI mentions “Product X” in one answer, it can easily continue talking about it in follow-ups without confusion. A good practice is to include brief re-introductions of key entities when transitioning topics (much as a well-written article might do). This mirrors what the AI will do – it often rephrases or reintroduces context for itself as the conversation goes on. Provide summary and recap sections. Because an AI will keep earlier context in mind, if your content includes a short summary or highlights, it might preferentially use that summary when the user drills down. For instance, imagine a user asks “Tell me about company ABC.” The AI might pull a summary from ABC’s “About Us” page. If the next question is “What were their revenues last year?” – the AI might recall a specific figure from earlier content or it might quickly scan for a number in its stored context. If your content had a quick facts section (“Founded: 2010, Revenue 2024: $50M, Employees: 200”), the AI can answer directly with that data. If not, it might generate an approximation or skip it. Essentially, having structured data or concise facts in your content helps the AI retrieve those facts in multi-turn conversations.

(Structured data markup can help as well – we’ll cover that in the technical chapter – and indeed Google’s AI overview has been said to leverage structured data where available.) Conversational tone and FAQ format can be beneficial. This will be discussed more in Chapter 9, but in brief: content that is already in a conversational Q&A style is naturally aligned with how users interact with chatbots. Many businesses are now adding FAQ sections or conversational snippets to pages (e.g., “Q: What does this product do? A: It helps you…”) not just for traditional SEO (featured snippets) but for AI. If a user asks an AI “What does [Your Product] do?”, the AI might directly use the Q&A from your site if it’s clearly written, rather than synthesizing an answer from scratch. Moreover, in a multi-turn exchange, if the AI gave a general answer initially, and the user asks a more specific question, the AI might look for a specific Q from an FAQ that matches. Embedding likely user questions into your content (and answering them) is a strategy (often called answer engine optimization (AEO) in the SEO community) that overlaps heavily with GEO. You want to be the source of those bite-sized answers the AI delivers. Let’s illustrate how multi-turn memory can play out with a hypothetical scenario: A user is planning a vacation and interacting with a travel AI assistant. The user says, “I’m interested in visiting Greece in September.” The AI might give an overview of Greece in September, perhaps noting weather, events, etc., citing sources like travel blogs or Greek tourism sites. Next, the user asks, “What are some must-see historical sites there?” – because the AI remembers we’re talking about Greece, it doesn’t need the user to repeat “in Greece”.

It will list sites like the Acropolis, Delphi, etc., maybe quoting a site about Greek historical attractions. Then the user says, “How about on the islands? I’m thinking Crete or Rhodes.” Now the AI needs to recall that we are still on the topic of historical sites and Greece, and specifically islands. It might then give an answer about the Palace of Knossos in Crete and the Colossus site on Rhodes, for example, pulling from those specific island tourism guides. In doing so, the AI might have retrieved information from different pages for each turn, but it keeps the conversation flow. From an optimization standpoint, if you are a travel marketer wanting your content in that mix, you’d want to have pages like “Top Historical Sites in Greece,” and maybe separate ones like “Top 10 Things to Do in Crete” that mention Knossos. The AI could use the first page for the mainland answer and the second for the island follow-up. If your Crete page has a section titled “Historic Attractions in Crete” with Knossos clearly described, the AI can more readily pull that for the user’s follow-up question about islands. On the other hand, if your info is scattered or under generic titles like “All About Crete” (where history is buried under beaches and food info), the AI might miss it or not find it quickly enough in context. Another aspect of memory is persistent user preferences or data. Some advanced AI systems (and likely future personal AI assistants) could remember user-specific info across sessions (with permission). For instance, a user might tell an AI “I have gluten allergy” in one turn, and later on, while asking about recipes or restaurant recommendations, the AI will keep that in mind and filter answers.

As a marketer, consider how your content might be parsed in light of such personalized context. If you run a restaurant and have a menu page, clearly labeling which items are gluten-free or vegan (with text the AI can read, not just icons) will be important so that an AI assistant can say “Yes, this restaurant has 5 gluten-free entrees” in a conversation. Essentially, clarity in content helps AI not just generally but in delivering personalized answers matching the user’s context. It’s also useful to understand the limits of AI memory. While LLMs can carry a lot of context, they do have finite windows. For example, ChatGPT with GPT-4 8K can handle roughly what’s in a few pages of text; with 32K, maybe a small ebook’s worth. Claude with 100K can handle huge texts, but even that has limits (about 75k words as noted). If a conversation goes on and on, older parts of the dialogue might get dropped or summarized to stay within limits. So, if a user has a very lengthy interaction (say, 100+ turns), the AI might not perfectly recall details from the very beginning unless it was explicitly reinforced or repeated. That’s one reason why reinforcing key points in your content is helpful – if an AI saw it multiple times or in summary form, it’s more likely to stick in the conversation. The Anthropic example showed that the AI could find a very specific changed line in Gatsby because it had the whole text in context. But not every AI will load an entire page; many times they only use a snippet that looked relevant. If further questions require more detail from that page, the AI might go back and fetch more.

Ensuring your page is easily navigable (clear subheadings, jump links, etc., which an AI can use similarly to how a human would) could facilitate the AI retrieving the needed context again. Multi-turn SEO (or GEO) is still an emerging idea, but it boils down to this: Think about conversational workflows. In the past, you might have thought of search queries in isolation (“user searches X, lands on my page, leaves or converts”). Now, think in terms of dialogue: “User asks broad Q (AI gives overview, maybe mentions my brand), user asks specific Q (AI pulls a specific fact, maybe from my site), user asks comparative Q (AI might use data from me and competitor side by side), user then decides next step.” In that chain, you want your information to be present and accurate at each relevant step. This could mean having a mix of content: broad explainers for the top-of-funnel overview answers, detailed specs or data for the mid-funnel detailed questions, and perhaps even user-generated content or reviews (if you host those) for questions about experiences or opinions (which AI might surface e.g. “what do people say about Product X’s battery life?”). One more point: context extends to user context, not just conversation context. LLM-powered systems could use contextual signals like location, time, or user profile to tweak answers. For instance, if someone asks an AI voice assistant “What should I have for dinner?” the AI might consider it knows the user is vegan and it’s 5 PM on a weekday – it might answer differently for that user than for another. While this strays into personalization, it’s related to context memory because the AI could remember preferences. Marketers should be mindful of providing content that can feed into these contextual pivots.

If you have schema markup for your restaurant that indicates “vegan options available” or if your recipe site has tags for “quick weeknight recipe,” those pieces of data could influence whether the AI picks your content when the user’s context is known (e.g., weeknight + vegan filter). In short, structuring content for various contexts (dietary, seasonal, regional, etc.) can pay off in a world of personalized AI answers. To sum up, LLM memory transforms search into a conversation. This rewards content that is structured, clear, and modular enough to be used in a piecemeal yet coherent way. It also opens opportunities for guiding users down a journey through content via the AI. Each turn is like a query, and your content should ideally be the answer to one of those queries. When optimizing now, ask yourself: If I were a chat AI, what follow-up questions might a user ask after this, and do I have content that answers those? This aligns with strategies like intent mapping and content clustering that SEO specialists already use (anticipating user follow-up questions and covering them). Now, those follow-ups might happen with the AI as the intermediary, but the logic remains: comprehensive, well-organized content wins.

4.5 Ranking vs. Relevance in LLM Outputs

In the era of classic search engines, success was largely about ranking – could you get your page to rank #1 for a target query, or at least on the coveted first page of results? The battle for the top spot drove the entire SEO industry. In the emerging era of AI-generated answers, the game shifts to relevance and representation – ensuring your content is chosen by the AI as part of its answer, even if your site itself isn’t shown as a traditional “blue link.” This doesn’t mean traditional ranking factors are irrelevant (Google and Bing still have their algorithms feeding into the AI), but the paradigm of how content is delivered to users is changing. AI search is about answers, not links. As a Google article succinctly put it, in generative AI search, the system isn’t retrieving a whole page and showing it; it’s building a new answer based on what it understands . The AI might pull one sentence from your site, another from someone else’s, and then stitch them together in a coherent paragraph (often paraphrasing along the way). For the user, this is convenient – they ask a question and get an immediate, consolidated answer. But for content creators, it raises the question: How do I get the AI to “pick” my content as part of that answer? We can think of this as the content being ranked internally by the AI for relevance, even if no explicit ranking is shown to the user. One key is semantic clarity and structured information. LLMs interpret web content differently from search engine crawlers. They ingest the full text and analyze the relationships and meanings, rather than just looking at meta tags or link popularity. They pay attention to things like the order of information, headings and subheadings that denote hierarchy, and formatting cues (bullet points, tables, bold highlights) that signal important points. If your content is well-structured and clearly written, the LLM can understand it better, which increases the chance it will use it in an answer. Think of headings as signposts for the AI – a clear H2 like “Benefits of Solar Panels for Homeowners” tells the model that the following text likely contains a direct answer if a user asks “What are the benefits of solar panels for a homeowner?”. If instead your page has a clever or vague heading (e.g. “Shining Bright!” for that section), the AI has to infer what that section is about. It might still figure it out from the text, but you’ve added friction. As Carolyn Shelby writes in Search Engine Journal , poorly structured content – even if keyword-rich and marked up with schema – can fail to show up in AI summaries, while a clear, well-formatted piece without a single line of schema can get cited or paraphrased directly . In other words, content architecture beats metadata hacks in the AI world. That’s not to say schema isn’t useful (it can help, and Google has confirmed their LLMs do take structured data into account), but if you had to prioritize: make the core content extremely clear and skimmable. An AI is essentially a super-fast reader – it should be able to glance through your content and quickly grasp the key points. If your page has one H1 and 20 H2s all named something quirky, it might “confuse” the model’s understanding of what’s important. Logical nesting of H1 > H2 > H3 (with meaningful titles) essentially provides an outline to any reader, human or AI. Use that to your advantage. Direct answers and snippets. Just as featured snippets in Google were often drawn from concise answer boxes in content (like a definition in a single sentence, or a numbered list of steps), AI answers often prefer content that is formatted for easy extraction . Lists, steps, tables, and FAQs are golden ( ^[4] ). For example, if the query is “steps to change a tire,” an AI will very likely present a step-by-step answer. If your article “How to Change a Tire” has a nice ordered list of steps 1 through 5, there’s a good chance the AI might use your list (perhaps reworded) in its answer, possibly even citing you. If your article instead is a long narrative with the steps buried in paragraphs, the AI might still glean the steps but could just as easily use another site’s list instead. We see this with tools like Bing Chat – it often pulls bulleted or numbered lists from sites to give to the user (with little [source] annotations). Perplexity, which always cites sources, tends to pull concise statements. If someone asks a medical question, Perplexity might grab a one-line definition from Mayo Clinic or WebMD rather than a verbose explanation from elsewhere, because it’s easier to drop that one line into an answer. So, make your key points concise and standalone . This doesn’t mean oversimplify everything; it means consider adding summary sentences or bullet points that encapsulate the detailed text. A good practice is to front-load key insights in your content. Don’t bury the lede. LLMs, like rushed readers, often prioritize what comes first in a section or document. If the first sentence of your intro is a crisp summary of the answer, the AI might use that and move on, whereas if you only reveal the answer in the conclusion, the AI might have already compiled an answer from other sources by then. Another concept emerging is the “AI citation economy” ( ^[5] ). When AI summaries (like Google’s SGE or Bing) do cite sources, being one of those sources can drive some traffic and certainly visibility. There’s anecdotal evidence that being cited in SGE can result in clicks if users want to “learn more” and trust the snippet they saw. Bing’s citations [numbers] are clickable and some users do click them. So how to get cited? Based on observation and some studies (like the BrightEdge analysis of SGE vs Perplexity citations), authoritative domains have an edge , and content format matters. Authoritative doesn’t just mean high Domain Authority; it can also mean niche authority. For instance, a well-structured blog post from a lesser-known site can still get cited if it directly answers the question better than a higher-authority site. That said, sites like Wikipedia, Britannica, official government sites, etc., are heavily cited by AI because they are factual and straightforward. If you’re outranked by such sites in regular search, you’ll likely also be “out-cited” by them in AI answers. The strategy here is to find the questions where you have a unique value or perspective that the generic sources don’t, and ensure your answer is crystal-clear. For example, maybe no Wikipedia article gives a step-by-step of a specific software troubleshooting that your site does – then your steps might get picked by the AI. Or your e-commerce site might have very specific data on product dimensions that general sites don’t list, so an AI might cite you as the source for “weight: 1.2 kg” if a user specifically asks about that. The granularity and uniqueness of information you provide can make you the go-to source for certain details. We should also address overlap and differences between optimizing for classic SEO vs. LLM SEO (LLMO) . There’s overlap: things like clear headings, good content, authoritative backing – these were always SEO best practices and remain so. But gaps exist: for example, link-building, a cornerstone of SEO, might not directly translate to AI answer optimization. An LLM doesn’t care how many backlinks your page has when it’s generating an answer (though the retrieval algorithm that selects your page might use PageRank or similar, indirectly making backlinks still relevant). But the AI model itself isn’t making a judgment like “this site has lots of links, so its content must be good.” It’s judging content quality more intrinsically – coherence, completeness, readability. This suggests that on-page content quality matters even more , whereas off-page signals might influence whether you get retrieved in the first place (since search indexes and authority still feed into what content the AI sees). Google’s generative search, in particular, seems to often pull from pages that were already ranking in the top results (no surprise since it’s built on Google Search). BrightEdge’s study noted a lot of overlap in which domains get cited by SGE and by Perplexity. Big hitters like Wikipedia and official sites frequently appear. But also interesting is that different AI systems have different citation patterns – Perplexity, for instance, might favor certain tech forums or Reddit for some queries, whereas SGE might stick to more formal sources. Knowing these tendencies can inform strategy: e.g., if Reddit is often cited for certain tech questions and you run a tech company, maybe it’s worth engaging in those communities or providing expert answers there (so that when AI pulls from Reddit threads, your answer could be included). That ventures into off-page tactics, but it’s an example of thinking beyond your own site. Semantic relevance vs. exact keywords. We touched on this with retrieval, but it’s worth reinforcing: meaning is king for LLMs. If your content semantically answers the question, the AI can use it even if wording differs. However, when it has many choices, it may lean to content that more literally matches the question (since it’s “safer”). For instance, a user asks, “How can I improve my website’s accessibility for visually impaired users?” Suppose you have an article titled “Making Websites Screen-Reader Friendly” – that’s on-topic but not a word-for-word match. Another site has an article “How to Improve Website Accessibility for Visually Impaired Users” – basically the query terms as a title. If all else is equal, the AI might pick text from the latter because it’s an obvious direct match. This echoes traditional SEO advice: align with user language. In the SEJ example, the author’s article about “AI search” didn’t show up for an LLM query about “LLMs and schema” because it never explicitly said “LLM,” even though it was relevant ( ^[6] ) ( ^[7] ). The model had plenty of other content with the exact term, so it used those. The lesson: don’t shy away from using the same terminology your audience uses, even as you focus on depth and quality. LLMs are smart but when composing an answer, they might play it safe by quoting content that literally contains the asked-for terms ( ^[3] ). Finally, consider user engagement signals in an AI context . Traditional Google ranking has long debated using pogo-sticking or time-on-page as signals (with mixed evidence). In an AI answer scenario, if the user is satisfied, they might not click anything at all. If they’re not, they might click one of the cited sources or ask a follow-up. It’s conceivable that if users frequently click a particular citation after seeing an AI answer, that might be a sign that the citation had more to offer or the AI answer was lacking detail from that source. AI providers could use such feedback to adjust which sources are chosen or how answers are formulated. We don’t know for sure yet, but user behavior in interacting with AI results could indirectly affect which sources get favored over time . For marketers, this means if you do get cited, make sure the page the user lands on is high quality and answers what the AI snippet couldn’t (encourage the user to stay and explore). If the AI summary only gave a teaser and users click through to your site for full info, that’s great. But if the AI gave almost everything and the user has no reason to click, you got visibility but no visit. In such cases, think about content depth : maybe providing some unique value (interactive element, tool, community, etc.) beyond the text that the AI used. That can entice users to learn more on your site even after a good AI answer. In conclusion, **the focus shifts from ranking to being the reference the AI trusts.** In GEO, your content’s structure, clarity, and authority determine if the AI will choose your snippet in its synthesized answer. Clean, well-segmented content is now a kind of AI ranking factor ( ^[5] ). Success is measured by whether the AI includes your brand or content in answers (and ideally cites it) – even if the user never sees a traditional search results page. The overlap with traditional SEO is substantial – good content is good content – but the way that content is evaluated and utilized by AI introduces new nuances. By aligning with how LLMs parse and generate information, you increase your chances of remaining visible and relevant in a future where answers, not links, are the immediate output of a search. In this chapter, we explored the inner workings of LLMs from a marketer’s perspective: how they learn and generate text, how they handle new information and context, why they sometimes err, and how they incorporate content into answers. The recurring theme is that many classic SEO best practices (quality content, structured pages, understanding user intent) are not only still valid – they’re essential for GEO. At the same time, we must adapt to the nuances of AI-driven search: optimizing for an answer engine that writes summaries and engages in dialogue, rather than a static list of ranked links. As we move forward, the subsequent chapters will build on this foundation. Chapter 5 will dive into ChatGPT and generative Q&A specifically – including how ChatGPT differs from search and how one might optimize for it. Chapter 6 will explore Google’s approach (SGE, Bard/Gemini) and what that means for content creators. And so on through technical SEO adaptations, off-page signals in an AI world, prompt strategies, and measurement. Understanding how LLMs work is the first step; next, we’ll translate that understanding into actionable SEO and content tactics for the AI era.

References

^[1] Poloclub.Github.Io Article - Poloclub.Github.Io URL: https://poloclub.github.io/transformer-explainer

^[2] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Retrieval-augmented_generation

^[3] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[4] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[5] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[6] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[7] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

Chapter 5: OpenAI’s ChatGPT and the Rise of Generative Q&A

5.1 ChatGPT’s Introduction: The Breakthrough Chatbot in Mainstream AI Q&A

In late 2022, OpenAI’s ChatGPT burst onto the scene as a conversational AI that could answer questions and perform tasks through a natural dialogue interface. Its impact was immediate and massive. Within just two months of launch, ChatGPT’s user base skyrocketed to an estimated 100 million monthly active users – making it the fastest-growing consumer application in history at that time ( ^[1] ). By early 2025, ChatGPT had become one of the top 10 most visited websites in the world, attracting roughly 4.8 billion visits per month ( ^[2] ). To put that in perspective, nearly half of consumers (48%) in one late-2024 survey reported using ChatGPT or a similar AI tool in the past week alone ( ^[2] ). Such rapid adoption indicates that ChatGPT introduced generative AI to mainstream audiences on an unprecedented scale. ChatGPT as an “Answer Engine.” Unlike traditional search engines that provide a list of links, ChatGPT delivers a single, synthesized answer or solution in response to a natural-language prompt. Users have embraced this answer engine style for a staggering range of tasks. For example, software developers and students turn to ChatGPT for coding help or debugging advice instead of searching forums; in fact, the Q&A site Stack Overflow saw significant traffic declines (a 14% drop in one month) as coders opted to “just get an answer” from ChatGPT rather than browsing forum threads ( ^[3] ) ( ^[4] ). People also use ChatGPT for content creation (drafting emails, essays, marketing copy), for learning about complex topics through simple explanations, and even for personal tasks like brainstorming gift ideas or planning trips. A late-2024 retail survey found that 51% of shoppers had tried ChatGPT or similar generative AI tools in their daily lives (up from 29% a year prior) ( ^[5] ), using them for product research, personalized buying guides, and recipe discovery. In short, ChatGPT’s ability to engage in plain English (or any language) conversation and provide direct answers has made it a go-to digital assistant for millions of users. From Novelty to Ubiquity. ChatGPT’s rise also benefited from massive public curiosity and media attention. It was often described as “AI that can answer anything” , drawing users who tested it on everything from trivial questions to professional work. By mid-2023, businesses began exploring ChatGPT’s potential, and by 2025 about 28% of U.S. workers reported using ChatGPT in their jobs (up from just 8% in early 2023) ( ^[6] ) ( ^[7] ). The user demographic skewed young and educated at first – a Pew Research survey in early 2025 found 58% of Americans under 30 had used ChatGPT (compared to 33% of those 50 and older) ( ^[8] ) ( ^[9] ) – but awareness and usage have grown across all groups. Notably, ChatGPT’s launch prompted a public discourse about AI’s capabilities and risks (e.g. schools worried about cheating, publishers about content scraping), yet this did little to dampen enthusiasm. OpenAI capitalized on the momentum by continually improving the model (releasing GPT-4 in 2023) and introducing new features (like image/voice input and third-party plugins, discussed below). Each update expanded what ChatGPT could do, further entrenching it as a versatile digital assistant. In summary, ChatGPT was the breakthrough that familiarized the broader public with AI-powered Q&A. It demonstrated, at scale, that many search or help tasks could be accomplished through a conversation with an AI instead of a traditional search query . This paradigm shift – from typing keywords into Google to asking a question to an AI – has profound implications. For users, it offers convenience and efficiency; for content creators and marketers, it foreshadows a new landscape (Generative Q&A) in which providing information or solutions without a click to one’s website becomes the norm. In the following sections, we delve into how ChatGPT’s usage differs from traditional search and what that means for those seeking to optimize content in this new era.

5.2 ChatGPT vs. Traditional Search: A New User Behavior

ChatGPT introduced a fundamentally different user experience compared to traditional search engines like Google. Instead of entering terse keywords and browsing a list of blue links, users pose complete questions or requests in natural language – for example, “How can I improve my website’s SEO?” – and receive a single, coherent answer composed by the AI. This difference has led to distinct user behaviors and expectations: Conversational Queries: Users can ask ChatGPT questions in a straightforward, conversational manner, including follow-up questions to probe deeper or clarify – much like having a dialogue with an expert. By contrast, traditional search often requires iterative keyword tweaking and filtering through multiple results to piece together an answer. With ChatGPT, the first result is often the final result , since the AI crafts a synthesized response drawing from its vast training knowledge. One Answer vs. Many Links: Perhaps the biggest shift is that ChatGPT usually provides one answer (occasionally with multiple suggestions or options within it), whereas Google provides dozens of hyperlinks for the user to choose from. For users, this one-stop answer can be appealing – it’s fast and requires minimal effort to get information. In a controlled comparative study, participants using ChatGPT completed information-finding tasks significantly faster than those using Google Search, spending minutes less on average per task ( [10] ).

They also issued fewer or comparable follow-up queries, but in a more conversational style (longer, detailed questions rather than terse keywords) ( [11] ). Despite having fewer sources to cross-check, users in that study rated ChatGPT’s information quality higher and found the experience more useful and satisfying than a standard web search ( [12] ). This highlights how a well-written, consolidated answer can trump an array of search results in perceived value. Interactive Refinement: With search engines, refining a query means trying a new search or clicking on advanced filters. With ChatGPT, users can simply ask the AI to adjust the answer: “What if my budget is only $500?” or “Can you clarify that last point?” . The AI remembers the context of the conversation and tailors its next response. This multi-turn interactivity makes information retrieval feel like a conversation , reducing the friction of starting new searches from scratch. It also enables users to explore a topic in-depth without ever leaving the chat interface. These differences have major implications for content creators and website owners . In the ChatGPT model, the AI often satisfies the user’s information need without the user ever clicking through to an external source . The traditional search paradigm at least presented users with source links (and often a portion of content via featured snippets) that a user could click for full details.

With ChatGPT’s default mode, however, the user might get a perfectly sufficient answer composed from various sources, and no attribution is shown at all . For example, where a Google search for a programming error might show a snippet from Stack Overflow and invite a click for more context, ChatGPT might explain and solve the error right within the chat , negating the need to visit Stack Overflow. It’s no wonder that Stack Overflow’s traffic declined sharply when ChatGPT usage surged ( [3] ) ( [4] ). As one analyst noted, “ChatGPT users miss out on the debate and just get an answer, which can seem quicker and more efficient” , whereas on a forum multiple answers would be posted and voted on ( [13] ). This encapsulates the convenience that draws users to ChatGPT – and the dilemma it poses for content platforms that traditionally banked on users clicking through for nuanced discussions. User Trust and Verification. Another difference is how users perceive the information. Google results inherently encourage cross-verification – a user might open two or three links to compare answers, aware that each site has its own perspective. ChatGPT, by contrast, delivers information with a single authoritative voice, which some users may take at face value.

Studies have found that users often trust ChatGPT’s answers and enjoy the experience – even rating its answers’ quality quite high ( [12] ) – but there’s a risk here: if the AI’s answer is incorrect (a known issue, as large language models can “hallucinate” false facts), the user might not have immediate cues to doubt it, especially since source references are usually absent. Traditional search at least displays source names (e.g. a well-known news site vs. an unknown blog), which help users gauge credibility. ChatGPT strips away those cues in its default answer presentation. This has led to instances of misinformation spreading quietly – for example, users relying on confidently stated but inaccurate medical or financial advice from ChatGPT, which they might have caught if they had seen conflicting sources via search. As we will explore later, ChatGPT’s lack of transparent sourcing is a double-edged sword: it provides a clean, simple answer experience but complicates the user’s ability to verify information. Shifting Search Behavior – Early Trends. While ChatGPT has not replaced traditional search engines at large, there are signs of a shift in certain segments and query types. One survey in late 2024 found that around 3.8% of consumers were already using AI-powered tools like ChatGPT or Anthropic’s Claude in place of search engines for some queries ( [14] ).

That’s a small fraction compared to the ~83% who still use Google primarily ( [15] ), but notable given AI search tools were only recently available. This adoption was higher among younger users and those frustrated with traditional search quality (e.g. Google’s ad-ridden results and CAPTCHAs) ( [16] ) ( [17] ). Indeed, mounting frustration with Google’s experience – 66% of users in one poll complained of too many ads, and ~45% noted issues with Google’s own AI-generated snippets ( [16] ) – is likely driving some to seek alternatives like ChatGPT ( [17] ). As ChatGPT and similar tools improve and become more integrated (for instance, via browser extensions or mobile apps), we can expect the percentage of “AI-search” users to grow. Google itself has recognized this change, fast-tracking its own conversational AI features (covered in Chapter 6) to retain users who might otherwise defect to ChatGPT for quick answers. From a marketer’s perspective , the rise of ChatGPT means that the playing field of SEO is expanding beyond traditional search engines . We now have to consider how our audience’s questions might be answered by an AI agent that doesn’t necessarily lead users to our website. In practical terms, this raises questions like: How do we get our information into the answers that ChatGPT provides? and How do we maintain brand visibility when the “interface” is just an AI’s text response?

These are complex challenges that we will address in upcoming sections. But first, let’s look at how OpenAI has extended ChatGPT’s capabilities with real-time data access – a development that creates both new opportunities and new optimization considerations for content owners.

5.3 Plugins and Browsing: Extending ChatGPT’s Reach to the Live Web

One limitation of ChatGPT’s initial release was its knowledge cut-off date (originally September 2021 for GPT-4). By design, the base ChatGPT model did not have awareness of events or content created after that cut-off, meaning it couldn’t answer questions about recent news, up-to-date stock prices, this week’s weather, and so on. To address this and expand ChatGPT’s functionality, OpenAI introduced two major features in 2023: third-party plugins and a web browsing mode . These additions effectively gave ChatGPT access to real-time information and specific external services, transforming it from a static Q&A model into a more dynamic platform. Web Browsing Mode (via Bing). In mid-2023, OpenAI rolled out a feature called “Browse with Bing” for ChatGPT (initially to ChatGPT Plus subscribers). This allowed the chatbot to perform live web searches and read content from the internet when a user’s query required up-to-date info or more data than the AI’s training provided. OpenAI built this capability in partnership with Microsoft, leveraging the Bing search index and API behind the scenes ( ^[18] ) ( ^[19] ). Essentially, when browsing mode is enabled, ChatGPT formulates a search query, retrieves results from Bing, clicks on relevant pages, and can quote or summarize those page contents in its answer. Notably, OpenAI designed the browsing feature to include direct citations (links) to the sources it pulled information from ( ^[20] ). In a public announcement, OpenAI emphasized that ChatGPT with browsing could provide “current and authoritative information, complete with direct links to sources.” ( ^[20] ) This was a significant departure from ChatGPT’s default behavior of giving unsourced answers, and it mirrored the approach of other AI search engines (like Bing’s own chatbot and Google’s Bard) which cite sources for transparency. The browsing feature had a bit of a rocky launch. It was first introduced in beta around May 2023 ( ^[21] ), then temporarily disabled in July 2023 due to concerns that ChatGPT might display copyrighted or paywalled content without permission (essentially acting as an unintentional circumvention tool) ( ^[22] ). OpenAI reworked the feature and reintroduced Browse with Bing by the end of September 2023 ( ^[23] ) ( ^[22] ), this time with measures to respect content owners. The relaunched version allowed website publishers to control how ChatGPT’s web agent interacts with their site ( ^[24] ) ( ^[25] ) – for example, sites could use standard mechanisms like robots.txt or special meta tags to signal if the AI should or shouldn’t access their content. This gesture was aimed at alleviating publisher worries: OpenAI publicly stated it wanted to “do right by content owners” and would work on solutions so that the browsing AI respects rules and potentially paywalls ( ^[26] ) ( ^[25] ). From a content creator’s perspective, ChatGPT’s browsing mode is a double-edged sword. On one hand, it creates new opportunities for your content to be surfaced and credited in AI-generated answers. If a user asks ChatGPT a timely question (say, “What were the results of yesterday’s election?” or “What’s the latest iPhone review say?” ), ChatGPT may search the web and find a news article or blog – perhaps yours – and then include information from it with a link . This means your site could gain visibility whenever ChatGPT “chooses” it as a source. In effect, being referenced by ChatGPT in browsing mode is similar to being the featured snippet on Google, albeit with potentially even greater impact since the user might be fully relying on that single answer. If the user wants more detail or to verify, they have the direct link to click through. Indeed, marketers have noted that ranking well on Bing (which feeds ChatGPT) can lead to being cited in ChatGPT’s answers, making Bing SEO an important part of Generative Engine Optimization (GEO) ( ^[19] ) ( ^[27] ). We’ll discuss optimization shortly, but the key point is: strong organic presence in Bing’s index is now critical, since content not indexed (or deindexed due to a Bing penalty) won’t appear in ChatGPT’s web-enabled results ( ^[28] ) ( ^[29] ). Microsoft’s own product lead for Bing confirmed that “if you currently want to rank in ChatGPT Search, you need to be indexed by Microsoft Bing Search” , as ChatGPT’s web search draws from Bing’s index ( ^[29] ). This underscores that traditional SEO fundamentals (like submitting sitemaps to Bing, monitoring Bing Webmaster Tools, etc.) remain highly relevant in the age of ChatGPT. On the other hand, ChatGPT’s ability to directly provide answers from your content can still deprive you of a visit or user engagement. A user might get what they need from the snippet ChatGPT pulled and have little incentive to visit your page. At least with the link present, a motivated user can click through – and some will. For example, if ChatGPT summarizes a complex how-to article and cites the source, a user interested in nuances might click to read the full piece (much as they might click a featured snippet source for depth). There is also a branding benefit: having your site’s name or URL shown as the source lends credibility to the info and creates awareness ( even if the user doesn’t click immediately ). This brand exposure aspect is something to strive for, since ChatGPT’s default mode (without browsing) does not give any explicit credit. Third-Party Plugins and Integrations. Alongside web browsing, OpenAI debuted a plugin ecosystem for ChatGPT in 2023 that enabled the AI to interact with specific services and data sources beyond its core training. The first wave of plugins included integrations with well-known platforms: for instance, Expedia and Kayak (travel search – enabling ChatGPT to help plan trips and find flights/hotels), OpenTable (restaurant reservations), Instacart (grocery orders), Klarna (shopping price comparisons), WolframAlpha (advanced math and computational queries), Zapier (to perform actions across dozens of productivity apps), and others ( ^[30] ) ( ^[31] ). With plugins, ChatGPT became not just an information source but a tool-using agent . A user could ask, “ChatGPT, book me a table for two in New York tomorrow night” and with the OpenTable plugin the AI could search for restaurants and actually return reservation options. Or one could ask to “analyze this dataset” and the Code Interpreter (later renamed Advanced Data Analysis) plugin would allow file uploads and perform Python calculations. For content publishers and businesses, plugins offered a novel way to integrate with ChatGPT’s flow . A retailer, for example, could create a plugin such that when a user asks “I need a new laptop under $1000,” ChatGPT might use the retailer’s plugin to fetch live product data and possibly recommend items from that store. Some early adopters like Shopify experimented with this concept, allowing ChatGPT to pull product information from Shopify stores via a plugin. In theory, a plugin could drive highly qualified traffic or conversions by inserting your service at the point of inquiry (the user might complete a purchase or sign-up via the plugin without ever “visiting” the website in a traditional sense). However, it’s worth noting a recent development: in early 2024, OpenAI decided to sunset the original plugin system in favor of a new approach to extensions called ChatGPT “Custom GPTs” (or GPT-initiated actions) . By April 2024, the plugin store was phased out ( ^[32] ) ( ^[33] ) and users could no longer initiate new conversations with those standalone plugins. OpenAI replaced this with a mechanism allowing developers or users to create custom versions of ChatGPT with specific knowledge or tool integrations – essentially merging the idea of plugins into specialized AI assistants (often just called GPTs). For example, instead of enabling a “Shopify plugin” in a conversation, one might use a custom GPT that is designed for shopping and already knows how to use the Shopify API internally. From a user perspective, the experience is similar (ChatGPT can fetch real-time info or perform tasks beyond its base model), but the management is different behind the scenes. The key takeaway for our purposes is that ChatGPT still has the capability to use tools and fetch live data – whether via the legacy plugin model or the new GPT-based model – and this capability is likely to expand. OpenAI’s move indicates they want tighter control and smoother integration of these tools, perhaps to improve usability and safety. As marketers, we should follow these developments; if custom GPTs allow businesses to publish AI “agents” or integrate data, we’ll want to be there. The specific mechanisms may change (plugin vs GPT), but the principle stands: content and services can be plugged into ChatGPT’s brain, which is a channel to reach users directly in their AI conversations. Ensuring AI Accessibility of Your Content. The advent of ChatGPT’s browsing and plugins means that our content might be consumed by AI in new ways. Here are some considerations to ensure your content is AI-accessible and AI-friendly : Allow Crawling by AI Agents: If you had previously blocked OpenAI’s crawler or other AI bots (perhaps out of concern for content misuse), reconsider that stance in light of the opportunities. OpenAI’s web crawler (originally identified as OAI-SearchBot and now as GPTBot ) obeys robots.txt . Some websites added rules in late 2022/early 2023 to disallow it (out of fear of being scraped for training data). If your site was among them, it’s time to lift those blocks if you want ChatGPT to utilize your content ( ^[34] ) ( ^[35] ). Updating your robots.txt to explicitly allow OpenAI’s bot is one step recommended for GEO (Generative Engine Optimization) ( ^[34] ) ( ^[35] ). Likewise, ensure you’re not inadvertently blocking Bing’s crawler, since ChatGPT relies on Bing – submitting an XML sitemap to Bing Webmaster Tools can help ensure all your pages are indexed there ( ^[36] ). In short, treat AI crawlers with the same importance as Google’s crawler: they are gateways to visibility in AI answers. Site Structure and Performance: ChatGPT’s browsing mode tries to retrieve answers quickly – it may not crawl deep or wait long on slow pages. A well-structured website (with clear hierarchy and internal links) can aid the AI in finding relevant info fast ( ^[37] ) ( ^[38] ). If your page that contains the answer is buried several clicks down or not linked cleanly, the AI might miss it. There is anecdotal evidence that complex navigation can lead ChatGPT’s browser to time out or skip pages in favor of easier targets ( ^[39] ). To mitigate this, follow good practices that also benefit human SEO: use a shallow link depth (important pages reachable in a few clicks), use descriptive URLs and page titles, and maintain a logical hierarchy of headings. Moreover, ensure your pages load reasonably fast and aren’t overly reliant on client-side scripts to render key content – the AI likely reads the raw HTML. If important information only appears after a user login or after running a script, ChatGPT won’t see it. Content Formatting for AI Consumption: Just as we format content for human readability and snippet optimization, we should format for AI readability. Use descriptive, hierarchical headings (H1, H2, H3, etc.) and bullet points or numbered lists when appropriate ( ^[40] ). This not only helps human readers scan, but also helps ChatGPT parse and extract the essence of your content. If a page cleanly answers “What is X?” in the first few sentences after a heading, the AI can quickly grab that. If the page rambles or buries the answer in fluff, the AI might overlook it or produce a less accurate summary. Comprehensive content is encouraged (cover the topic in depth so the AI has all it needs) but avoid unnecessary fluff or filler ( ^[41] ). One GEO expert put it this way: give search engines and GenAI “all the data they need in one place” ( ^[41] ) – be thorough but stay relevant. Including FAQs on your pages (common questions and answers) can also be very useful. An FAQ section essentially pre-formats likely user questions in a way that ChatGPT might directly reuse. The model might even directly quote an FAQ answer if it’s succinct and on-point. Meta-data and Schema: While it’s still unclear how much ChatGPT utilizes structured schema markup, using schema (like FAQPage, HowTo, Article, etc.) on your content can’t hurt and may help Bing or other search engines present your content in a way that the AI can identify key pieces (for example, an FAQ schema explicitly tells a crawler the Q&A pairs on the page). Additionally, proper meta titles and descriptions might influence what the AI sees as a summary of the page when deciding to click it. Ensuring your pages have clear, content-rich <title> tags and headers will align with the terms the AI might be searching for. In essence, ChatGPT’s leap to live data through browsing and plugins means SEO now extends to catering to an AI reader as well as human readers . The opportunity is that your content can reach users even when they don’t explicitly seek it out – the AI can bring your insights into the conversation proactively. The challenge is that this happens without the traditional branding and conversion funnel of a website visit. In the next section, we’ll discuss how to optimize your content strategy so that, when ChatGPT (or similar AI) is answering questions in your domain, it draws from your content or mentions your brand . This is the crux of Generative Engine Optimization: influencing the answers of AI “engines” to include and favor your content.

5.4 Optimizing Content for ChatGPT Visibility

How do you get your content and brand into ChatGPT’s answers ? This is the central question of Generative Engine Optimization (GEO). While the algorithms and ranking factors for AI answers are not as transparent or established as Google’s PageRank ecosystem, early research and experience suggest several strategies. The encouraging news is that many classic SEO best practices still apply – quality content, authority, and technical accessibility remain paramount – but there are also new tactics specific to AI. 1. Maintain Strong Traditional SEO Signals (They Still Matter). ChatGPT’s browsing mode and any future “search” functions rely on existing search engines and link structures to find content. As noted, Bing’s index is a primary source ( ^[18] ) ( ^[19] ). Therefore, optimize your site for both Google and Bing : ensure all your important pages are indexed on Bing, perhaps by using Bing Webmaster Tools and monitoring for any crawl errors or penalties. One SEO expert demonstrated that a site penalized (removed) from Bing’s index stopped appearing in ChatGPT’s answers, even though it ranked fine on Google ( ^[28] ). In practice, this means diversifying your SEO efforts – don’t neglect Bing-specific optimizations like utilizing Bing’s URL indexing API for rapid indexation of new content. Likewise, acquiring backlinks from authoritative sites can indirectly boost your content’s presence in AI answers, because those links help your content rank on search engines which the AI is using as a filter for quality. In short, the foundation of GEO is still SEO : if your content isn’t discoverable by search crawlers or isn’t deemed authoritative/relevant for a topic, ChatGPT is unlikely to use it. 2. Create Authoritative, Original Content That AI Will Trust. ChatGPT’s goal (when it sources external info) is to provide reliable, accurate answers . When deciding which content to draw from or mention, the AI (implicitly or explicitly) considers factors akin to credibility and relevance ( ^[42] ). In fact, when asked what factors it considers for referencing a website, ChatGPT itself listed credibility, relevance, accuracy, recency, and user engagement ( ^[42] ). Independent analyses by SEO experts align with this: they found that brand mentions and reputation play a big role, as do traditional authority signals like backlinks and positive user reviews ( ^[43] ). What does this mean for content creation? Demonstrate E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) in your content. These principles, championed by Google, help ensure your content is high-quality – and they likely influence AI as well ( ^[44] ) ( ^[45] ). Write with a confident, knowledgeable tone, cite credible sources for facts (so that if ChatGPT trained on your article, it sees that you yourself back up your info), and provide author bylines or bios that establish expertise. Incorporating first-hand experiences, case studies, or original research can make your content stand out as unique and trustworthy (since it’s not just rehashing what’s elsewhere) ( ^[46] ) ( ^[47] ). For instance, if you publish a proprietary study (“Brand X 2024 Industry Survey”) with interesting statistics, you not only earn backlinks and press (helping SEO), but you also supply ChatGPT with a fact it might rely on and attribute to your brand in an answer. A concrete example: an original statistic like “According to Acme Corp’s 2024 survey, 67% of marketers plan to increase AI budgets” could very well surface in ChatGPT’s answers to a question about marketing trends, with your brand name included – because the AI often incorporates the source when quoting a distinctive fact ( ^[13] ). High-value original content thus serves a dual role: it attracts human attention and it seeds the AI with knowledge that’s explicitly linked to you. Be Comprehensive and Specific. Aim to cover users’ questions in depth, addressing the many facets of a topic, but also ensure that specific common queries are directly answered within your content. If you have a comprehensive article on “Guide to SEO in 2025”, break it down into clear sections that answer likely sub-questions ( “What are the latest ranking factors in 2025?” , “How has AI changed SEO?” , etc.). This increases the chance that ChatGPT, in answering a granular question, will find the relevant snippet in your article to use. It’s been observed that GenAI tends to favor content that it can easily extract succinct answers from – so include succinct answers within your longer content . You might start a section with a two-sentence summary answer (for quick uptake by AI), followed by the detailed explanation. Think of these as mini “featured snippets” embedded in your content. Use FAQ and Q&A Formats. Many organizations are now adding FAQ sections or even entire Q&A pages addressing common questions in their domain. This is wise for GEO because it mirrors the way people interact with ChatGPT (they ask questions). A well-structured FAQ page with questions as headings and clear answers can be a goldmine for ChatGPT. For example, if you run a travel site, an FAQ like “Q: What is the best time to visit New Zealand? A: The best time is …” might be directly lifted into an AI answer to someone asking that question. Even if not lifted verbatim, it ensures the AI finds a concise, context-aware answer on your site to draw from. Keep Content Fresh and Updated. Recency is one of the factors ChatGPT considers for reference ( ^[42] ), especially in fast-changing domains. Regularly update your key pages with the latest information, statistics, or developments. If ChatGPT’s browsing fetches two articles – one from 2019 and one from 2024 – it will likely favor the newer as long as it’s relevant (and especially if the user’s query implies they want up-to-date info). Indicate dates on your content (e.g. “Updated March 2025”) so both users and AI know it’s current. A practical tip is to audit your high-performing content periodically and refresh any outdated facts or references; this not only keeps your SEO strong (Google likes freshness for many queries) but also increases the odds that ChatGPT will view your page as a timely source worth citing. 3. Optimize for Branded Visibility in AI Responses. Beyond just getting facts from your site into ChatGPT’s answers, many marketers are concerned with getting their brand names into the answers – i.e. being explicitly mentioned or recommended by the AI. This is akin to the holy grail of GEO: when someone asks ChatGPT for a product or service recommendation in your category, your brand is one of the ones it names . Achieving this requires a combination of on-site and off-site strategies: Build Brand Mentions and Reputation Across the Web. Large language models like GPT-4 learned from vast swathes of internet text. If your brand is frequently mentioned in authoritative contexts – news articles, industry reports, high-quality forums, etc. – the model is more likely to “know” about it and possibly include it in relevant answers. Conversely, a brand with little online presence or only self-published material might be overlooked. In a 2024 analysis of AI recommendations, frequent positive brand mentions and appearing in “top recommended” type articles were found to boost the likelihood of being suggested by ChatGPT ( ^[48] ) ( ^[43] ). For example, if a user asks ChatGPT for “best project management software”, the AI will draw on what it has read – articles titled “10 Best Project Management Tools” from sites like PC Magazine, CNET, etc., as well as general sentiment. If your software appears in many such lists and has good reviews, it stands a far better chance of being named by the AI. This means digital PR and content marketing beyond your own site are crucial : get your brand featured in roundups, encourage customers to leave positive reviews on third-party platforms (ChatGPT might have knowledge of aggregated ratings), and engage in thought leadership that gets cited. These off-page factors mirror traditional SEO (where they improve domain authority), but here they also directly influence the AI’s training data and real-time data. One notable trend is major publishers partnering with AI companies (OpenAI has struck content deals with the likes of AP, News Corp, Condé Nast, and others) to integrate their content into AI models ( ^[49] ) ( ^[50] ). If your company has the means, being included or mentioned in the content of such partners (e.g. a mention in The Wall Street Journal, now known to be licensed to OpenAI ( ^[51] ) ( ^[52] )) could mean your brand is explicitly accessible to ChatGPT’s knowledge base, possibly with attribution. Even without such partnerships, getting quoted in reputable sources (that are likely part of the AI training corpora) is beneficial. Leverage Your About and Brand Pages. Ensure your own site’s “About Us” page and product pages clearly state who you are, what you do, and what makes you notable. This sounds basic, but remember, if ChatGPT is trying to answer “What is [Your Brand] known for?” or incorporate a one-line description of your company in an answer, it will look for a succinct description. A well-crafted paragraph on your About page (e.g. “Acme Corp is a leading provider of analytics software for small businesses, known for its innovative AI-driven platform.”) could become the very text the AI outputs when asked about your brand. An SEO agency observed that optimizing the About page with comprehensive, PR-style info can help the brand appear in AI recommendation lists ( ^[53] ) ( ^[54] ). Essentially, treat some of this content as if you were providing the AI a cheat-sheet about your brand. Include your key products, awards, years of experience, etc., in plain language. If you have a Wikipedia page, that’s even better – those are often used by language models as definitive sources. Keeping your Wikipedia entry accurate and updated can indirectly feed AI models correct information about your brand. Embed Likely User Prompts in Your Content. (This is a nuanced tactic that we explore more in Chapter 12 on Prompt Optimization, but it’s worth mentioning here.) Think about the exact questions users might pose to ChatGPT that are relevant to your business. For instance, if you run a content marketing firm, users might ask ChatGPT, “How do I increase my blog’s traffic?” It would be wise to have a blog post on your site titled “How to Increase Your Blog’s Traffic – 10 Strategies” , containing a direct answer. Better yet, phrase key lines in a way that aligns with the question: “If someone asks how to increase blog traffic, the answer includes: improve SEO, create quality content, promote on social media, etc.” It sounds meta, but by phrasing content in a Q&A style or conversational tone, you make it easier for ChatGPT to identify that your content is an answer to that question . Some site owners are even including a short Q&A at the bottom of articles restating the main question and summary answer (almost like an AI-oriented FAQ). Caution: this should be done in a genuine, useful way – do not just stuff a bunch of predicted questions with no value add, or you veer into spam. But carefully embedding likely prompts can guide the AI. Essentially, you’re anticipating the prompt and ensuring your content aligns with it. 4. Utilize Monitoring Tools to Measure AI Visibility. Just as SEO managers track keyword rankings on Google, a new practice is emerging: tracking your “AI search presence” . How often and in what context does ChatGPT mention your brand or content? This is admittedly challenging, since ChatGPT’s answers are not publicly indexed like web pages. However, tools and creative methods are appearing. For example, some SEO platforms have begun offering features to query ChatGPT (and other AI like Bing Chat, Google’s SGE, etc.) at scale with a set of queries and detect which brands or sources are referenced. One such tool, ChatBeat , claims to show how often your brand appears in AI answers to key industry questions, providing an “AI Visibility Score” ( ^[55] ). Another, Brand Radar by Ahrefs, was mentioned as tracking brand mentions in ChatGPT and Perplexity responses ( ^[56] ). These tools typically work by using the AI’s API or a browser simulation to feed a large list of relevant questions (the kind of questions your target audience might ask) and then parsing the responses for mentions of your brand name or website. Using these, you could establish a baseline (e.g. “Our brand was mentioned in 5% of 1000 relevant AI queries this month”) and track improvement over time as you implement GEO strategies. If dedicated tools are out of reach, a simpler DIY approach is to periodically manually test ChatGPT with queries you think your customers might be asking. Vary the phrasing and see what answers come up: Does the AI ever mention your brand or cite your content? If competitors are being mentioned and you are not, analyze why. Perhaps they have a widely referenced whitepaper or were covered in press articles that the AI absorbed. This can inform your PR or content focus (you might realize, for instance, that “ChatGPT keeps mentioning Competitor X’s report on cloud security – we should produce research in that area or get ours more widely published”). Also try asking ChatGPT directly about your brand ( “What is Acme Corp?” ). While you must take AI responses with a grain of salt (they could be outdated or hallucinated), it’s enlightening to see what the model “knows” or assumes about your brand. If it returns incorrect info, that’s a signal you need to correct public narratives (perhaps updating your own site or external sources with the right info). If it knows very little, that suggests your brand’s online footprint in the AI’s training data is minimal. Additionally, monitor your web analytics for traffic that might be coming indirectly via ChatGPT. Pure ChatGPT (the default OpenAI interface) doesn’t pass referrals the way a browser does, so you won’t see “chat.openai.com” in your Google Analytics referring sites. However, if ChatGPT’s browsing mode sends traffic, it might show up with a Bing referrer (since the click technically comes through Bing). Or users who see your brand in an AI answer may later google you or navigate directly – so you might notice spikes in branded search or direct traffic after certain events. These are subtle and hard to tie definitively to ChatGPT, but be aware of the potential patterns. 5. Encourage Engagement and Loyalty Beyond the AI Interaction. If a user gets an answer from ChatGPT that involves your brand, how can you turn that into a lasting relationship? This is more about post-AI engagement strategy than about the AI itself, but it’s crucial. For example, suppose ChatGPT in browsing mode answers a question by summarizing your blog post and provides a link. To capitalize on this, make sure that if the user does click through, your page is welcoming, loads fast, and clearly provides additional value beyond what the summary gave. The user should feel, “Ah, there’s more context or useful detail here that the AI couldn’t include.” This could convert them from a one-time info seeker into a regular reader or customer. Ensure your branding is strong on the landing page and include gentle prompts like “Subscribe for more insights” or “Download our full guide” – because while ChatGPT might have answered their immediate question, you can offer deeper engagement that the AI cannot (such as interactive tools, community, or personalized services). In cases where ChatGPT mentions your product among others (say, “Product X, Y, and Z are popular options”), a user might ask the AI for more about just your product . This is another moment of truth: ChatGPT may then pull from reviews, your website, or other data to describe you. Prepare for this by seeding positive, factual information about your product wherever you can: on your site (robust product descriptions, FAQs), in app stores (for apps, ensure the description is informative), and in user-generated content (encourage satisfied customers to share their experiences on public forums or review sites, which could end up influencing AI output). The more consistent and favorable the information about your brand across the web, the better the odds that ChatGPT will “speak well” of you when asked. Finally, consider providing tools or content to the AI ecosystem. Some companies have begun experimenting with offering an open API or data source that AI agents can use (with proper controls). For example, a financial data provider might make a free API for basic stock info; ChatGPT via a plugin or future integration might tap into that, and in doing so, mention the provider’s name. While building a ChatGPT plugin in the original sense is no longer applicable (as of 2024, with plugins replaced by custom GPTs), the concept of integrating your service with AI assistants remains. Keep an eye on OpenAI’s platform announcements – if they allow businesses to publish custom GPT-powered agents or integrate data, it may be worthwhile to be an early adopter, just as some were with plugins. In summary, optimizing for ChatGPT requires a multi-faceted approach: technical SEO to ensure discoverability; content strategy to provide high-quality, answer-friendly material; off-site marketing to build your brand’s authority in the AI’s “eyes”; and monitoring to adjust your tactics . Many of these efforts overlap with good traditional SEO and PR practices. In fact, one 2025 industry study concluded that “the future of SEO is about mentions, authority, and AI relevance” , urging brands to secure visibility in AI-generated results by consistently publishing quality content and securing mentions in trusted sources ( ^[57] ). The overlap is clear: what’s good for human search often aligns with what’s good for AI – because both ultimately try to discern reliable, relevant information. However, even with perfect optimization, marketers must face the reality that ChatGPT often operates as a closed loop for the user. The final section of this chapter addresses those realities – the limitations of ChatGPT as a channel – and how to strategize around them.

5.5 Limitations and Considerations for Marketers

No discussion of ChatGPT’s impact would be complete without recognizing its limitations – particularly from the perspective of content publishers and marketers trying to derive value from it. While ChatGPT is an amazing answer engine for users, it introduces some fundamental challenges: Lack of Guaranteed Traffic or Attribution. Unlike a search engine results page where a user must click a link to get the full content, ChatGPT often provides the complete answer right in the chat. This means that even if your content informed that answer, you may receive no click, no site visit, and often no credit . ChatGPT’s default mode does not cite sources. The model was trained on countless web pages (likely including many of ours), but it regurgitates the knowledge in a synthesized form. From an ethical standpoint, this has raised debates – it’s essentially using publishers’ content to answer queries without driving traffic back. As marketers, we have to accept that **ChatGPT’s value to us is more about indirect exposure than direct referral traffic . If one of your key facts or insights is used in an answer, the user may never know your brand was behind it. They might walk away simply satisfied with “the answer from ChatGPT.” Browsing mode, as discussed, is an exception where sources are shown – but remember, not every user enables browsing or has access to it. Many users use the free ChatGPT which, as of 2025, still relies mostly on the internal GPT-4 (with knowledge to 2021 and some fine-tuning updates, but no live browsing unless they specifically use the Bing integration on Edge or similar). So for a large portion of interactions, ChatGPT is answering from its trained knowledge and will not spontaneously say “According to example.com…”. In fact, the model was originally designed not to quote large blocks verbatim from sources (to avoid copyright issues), which further reduces explicit attribution. Brand Awareness as the New ROI.** Given the above, companies should treat presence in ChatGPT answers as a form of brand awareness or thought leadership , rather than a direct lead generator. If your brand or product name gets mentioned by the AI in the course of answering a question, that is a win in itself – even if the user doesn’t click anything at that moment. It means your brand has been put into the user’s mind as part of the solution or information they were seeking. For example, if a user asks “What are some reputable email marketing platforms?” and ChatGPT responds, “You might consider platforms like MailChimp, Constant Contact, or SendinBlue, as they are known for high deliverability and robust features,” then each of those named companies just got a little boost of endorsement-style exposure . The user might not immediately visit those websites (they might even just ask ChatGPT a follow-up like “How do they compare?”), but the seed is planted. As a brand, you want to be on that list of AI-recommended options. So, one new key performance indicator (KPI) to consider is your “share of voice” in AI answers . Just as PR agencies measure how often a brand is mentioned in media coverage, marketers will increasingly measure how often their brand or content is echoed by AI. This is inherently hard to quantify (for now), but the earlier section on monitoring tools hints at emerging solutions. We might see the rise of an “AI mention share” metric, e.g., “In 100 common customer questions, our brand comes up 5 times while our competitor comes up 8 times – we need to close that gap.” Dealing with Misinformation and Hallucinations. Another limitation is the risk of ChatGPT providing incorrect or misleading information about your company or any topic. The model sometimes “hallucinates” – meaning it might assert something that isn’t true, often in a very confident tone. There have been cases where ChatGPT falsely stated that a certain company was involved in a scandal, or invented a feature that doesn’t exist in a product. From a brand safety perspective, this is concerning: an AI might unintentionally spread a damaging rumor or incorrect fact. Unlike search results, where misinformation (if present) likely came from a specific source you could contact or correct, with ChatGPT the “source” is a nebulous training corpus and the AI’s own pattern generation. What can marketers do? First, monitor . Regularly ask ChatGPT (and similar AI) about your brand and products to see if any false information comes up. If you find something problematic – say, it claims your software has a certain bug that it doesn’t, or confuses your brand with another – you can take steps to correct the record in the source material. That may involve publishing clarifications on your site, in press releases, or getting authoritative sites to cover the correction. The AI will eventually ingest those corrections via updates or browsing. OpenAI is also continually improving factual accuracy and has some feedback mechanisms. If it’s a severe issue (e.g., defamatory false info), you could use OpenAI’s feedback channels to report it, though there’s no guarantee of immediate fix. Over time, as AI companies license more verified databases and content partnerships (like the news partnerships OpenAI struck in 2023-2024 ( ^[51] ) ( ^[50] )), the hope is that the AI answers will rely more on vetted information and less on random forum chatter that might be wrong. Another strategy is to saturate the information space with correct info . If, for instance, ChatGPT is often unsure or wrong about the year your company was founded or who your CEO is, make sure that information is clearly stated on Wikipedia, your site, business directories, etc. The more the AI sees consistent facts, the less likely it will stray into fabrication. Essentially, claim your narrative in all the places the AI learns from. Ethical Considerations – Avoiding Black-Hat GEO. Whenever a new optimization frontier emerges, so do the temptations for black-hat tactics. We should address this: some may wonder, “Can I trick ChatGPT into promoting my site?” Perhaps by stuffing certain keywords or questions into forums that the AI will train on, or by using automated bots to create content that praises your brand excessively in the training corpus. Such tactics are highly discouraged – not only are they unethical (manipulating information ecosystems can lead to broader misinformation issues), but they’re likely to backfire. AI companies are becoming aware of prompt and data manipulation attempts. OpenAI, for example, might actively filter out content that looks spammy or overly promotional from its training data. Even if a black-hat GEO tactic worked briefly (say, you got the AI to quote a made-up “study” favoring your product), it could be corrected in the next model update, or your brand could suffer reputation damage if exposed. The safer, long-term strategy is the organic one : build real authority and let the AI pick that up naturally. User Journey Fragmentation. Marketers should also realize that AI assistants fragment the user journey. A person might get their initial answer from ChatGPT, then later in the day perform a related search on Google, then maybe ask Alexa (voice AI) a follow-up. The path to conversion may involve fewer direct website interactions and more AI-mediated steps. This complicates attribution – you might see a direct visit that converts, but the seed for that conversion was planted by ChatGPT hours before. It stresses the importance of integrated marketing : ensuring your messaging and presence is consistent across all channels (web, AI, social). If a user hears about you first from an AI, and then later sees a tweet from you or an ad, those should reinforce each other rather than present disjointed messages. Adapting Content Measurement. In the SEO world, we’re used to metrics like organic traffic, click-through rates, time on page, etc. With ChatGPT in the mix, some content won’t be consumed on your site at all – it might be consumed via AI summary. How do you measure success of a piece of content in that scenario? It challenges us to come up with new KPIs. One approach is content impact over content traffic . If your content was referenced by an AI (even without a click), it had an impact. This could be measured qualitatively (e.g., noting that “Our whitepaper was summarized by ChatGPT in an answer – meaning it’s reaching people beyond our site”). Some companies might track how their content is shared or paraphrased across the web (using tools like BuzzSumo or specialized LLM tracking tools ( ^[58] )). Over time, analytics suites may provide insights like “X number of AI queries led to a mention of your site” – but until then, we rely on proxy measures. One tangible metric is brand search volume . If ChatGPT mentions your brand and the user is intrigued, they might later search your brand name on Google (since that’s a common user behavior when they hear of something new). Thus an uptick in branded search queries or direct traffic could indirectly indicate AI-driven awareness. Keep an eye on Google Search Console data for your brand queries – if you see growth not attributable to other campaigns, maybe your GEO efforts are paying off. The Road Ahead: Dialogue and Governance. Finally, consider the broader environment: AI-generated answers are still a new phenomenon, and there’s ongoing discussion about how they should coexist with content creators’ interests. We see some moves toward governance and transparency . For instance, Bing Chat always includes citations for every paragraph it generates. Google’s SGE (Search Generative Experience) highlights key source links alongside its AI summary. OpenAI, by partnering with publishers, is implicitly acknowledging that sources deserve credit and that good content isn’t free – those deals allow content behind paywalls to be summarized with attribution and links back , ensuring the publisher can benefit ( ^[50] ). It’s possible that future versions of ChatGPT could integrate a more robust citation system generally (not just when browsing). OpenAI’s CEO has mused about connecting answers back to sources as a way to both improve accuracy and reward content creators. As a marketer, you’ll want to stay informed about such developments. If one day ChatGPT provides an “AI answer with sources” by default, that could re-introduce more of a traffic flow (as users might click the listed sources). In that case, all the GEO work you did means you’ll be those sources, ready to receive the clicks. In conclusion, engaging with ChatGPT as a marketer requires a mindset shift. Instead of purely driving users to our content, we are also letting our content go to the users , via the AI, and then finding ways to capture value from that arrangement. Success is measured not just in immediate clicks, but in mindshare and influence. If your brand becomes the one that AI assistants consistently mention or draw upon for certain topics, you have achieved a new kind of digital prominence – one that might not show up in yesterday’s web analytics, but will certainly manifest in real-world reputation and downstream benefits. As we continue through this eBook, we’ll explore parallel developments with other AI models and search engines (Chapters 6 and 7), and later dive into specific techniques (Chapter 12 on prompts, Chapter 13 on measurement) that complement what we’ve discussed here. For now, the rise of ChatGPT serves as the template for how search is evolving: from ten blue links to one synthesized answer; from SEO to GEO; and from optimizing for algorithms to optimizing for AI reasoning . Marketers and content creators who recognize these shifts early and adapt their strategies accordingly are positioning themselves to thrive in the generative AI era, turning what might seem like a threat – the answer-without-click paradigm – into an opportunity for brand leadership in a new kind of search ecosystem. Sources: Reuters – ChatGPT sets record for fastest-growing user base ( ^[1] ) ( ^[59] ) SearchEngineLand – Optimize content strategy for AI-powered SERPs (2025) ( ^[2] ) ( ^[14] ) ArXiv (Xu et al. 2023) – ChatGPT vs. Google Search study ( ^[10] ) ( ^[12] ) Gizmodo – Stack Overflow traffic drops as coders opt for ChatGPT ( ^[3] ) ( ^[4] ) Al Jazeera – ChatGPT can now browse the internet… ( ^[60] ) ( ^[20] ) Search Engine Roundtable – ChatGPT Search is powered by Bing’s index ( ^[18] ) ( ^[19] ) Forge and Smith – Generative Engine Optimization: SEO for ChatGPT ( ^[42] ) ( ^[36] ) SearchEngineLand – Brand visibility in AI (media partnerships) ( ^[50] ) ( ^[57] )

References

^[1] www.reuters.com - Reuters URL: https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01

^[2] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[3] Gizmodo.Com Article - Gizmodo.Com URL: https://gizmodo.com/stack-overflow-traffic-drops-as-coders-opt-for-chatgpt-1850427794

^[4] Gizmodo.Com Article - Gizmodo.Com URL: https://gizmodo.com/stack-overflow-traffic-drops-as-coders-opt-for-chatgpt-1850427794

^[5] www.mytotalretail.com - Mytotalretail.Com URL: https://www.mytotalretail.com/article/survey-says-shoppers-interested-in-genai-for-better-product-discovery

^[6] www.pewresearch.org - Pewresearch.Org URL: https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023

^[7] www.pewresearch.org - Pewresearch.Org URL: https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023

^[8] www.pewresearch.org - Pewresearch.Org URL: https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023

^[9] www.pewresearch.org - Pewresearch.Org URL: https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023

^[10] Arxiv.Org Article - Arxiv.Org URL: https://arxiv.org/pdf/2307.01135

^[11] Arxiv.Org Article - Arxiv.Org URL: https://arxiv.org/pdf/2307.01135

^[12] Arxiv.Org Article - Arxiv.Org URL: https://arxiv.org/pdf/2307.01135

^[13] Gizmodo.Com Article - Gizmodo.Com URL: https://gizmodo.com/stack-overflow-traffic-drops-as-coders-opt-for-chatgpt-1850427794

^[14] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[15] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[16] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[17] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[18] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/bing-powers-chatgpt-search-38345.html

^[19] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/bing-powers-chatgpt-search-38345.html

^[20] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[21] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[22] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[23] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[24] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[25] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

^[26] Community.Openai.Com Article - Community.Openai.Com URL: https://community.openai.com/t/why-did-my-web-browsing-option-disappear-topic-curation/286748?page=5

^[27] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/bing-powers-chatgpt-search-38345.html

^[28] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/bing-powers-chatgpt-search-38345.html

^[29] www.seroundtable.com - Seroundtable.Com URL: https://www.seroundtable.com/bing-powers-chatgpt-search-38345.html

^[30] The-Decoder.Com Article - The-Decoder.Com URL: https://the-decoder.com/chatgpt-plugins-openai

^[31] TechCrunch Article - TechCrunch URL: https://techcrunch.com/2023/03/23/openai-connects-chatgpt-to-the-internet

^[32] www.reddit.com - Reddit.Com URL: https://www.reddit.com/r/ChatGPT/comments/1b21c8k/plugins_will_be_replaced_by_gpts

^[33] Community.Openai.Com Article - Community.Openai.Com URL: https://community.openai.com/t/error-plugins-are-no-longer-supported/715523

^[34] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[35] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[36] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[37] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[38] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[39] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[40] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[41] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[42] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[43] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[44] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[45] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[46] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[47] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[48] Brand24.Com Article - Brand24.Com URL: https://brand24.com/blog/rank-brand-on-chatgpt

^[49] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[50] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[51] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[52] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[53] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[54] Forgeandsmith.Com Article - Forgeandsmith.Com URL: https://forgeandsmith.com/blog/generative-engine-optimization-geo-seo-chat-gpt

^[55] Brand24.Com Article - Brand24.Com URL: https://brand24.com/blog/rank-brand-on-chatgpt

^[56] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[57] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[58] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/optimize-content-strategy-ai-powered-serps-llms-451776

^[59] www.reuters.com - Reuters URL: https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01

^[60] www.aljazeera.com - Al Jazeera URL: https://www.aljazeera.com/news/2023/9/28/chatgpt-can-now-browse-the-internet-for-updated-information

6. Google's Generative Search: From Snippets to SGE and Gemini

In this chapter, we explore Google’s journey from traditional search engine optimization (SEO) toward generative search. We trace how early direct-answer features like featured snippets and voice responses (Answer Engine Optimization, or AEO) set the stage for today’s AI-driven Search Generative Experience (SGE). We also examine Google’s AI chatbot Bard and the upcoming Gemini model – key components in Google’s response to the large-language-model (LLM) revolution. Importantly, we’ll discuss how marketers can optimize for these AI-powered results using proven SEO best practices, and analyze the opportunities and threats generative search presents for various industries (including e-commerce, SaaS, and B2B).

6.1 Featured Snippets and the Roots of AEO (Answer Engine Optimization)

Featured snippets were Google’s first major step toward providing direct answers on the results page. Introduced around 2014, featured snippets are concise answer boxes that appear at the top of search results, pulling an excerpt from a relevant webpage. For example, a query like “What is a marketing funnel?” might show a boxed summary definition drawn from a marketing blog, above the usual list of blue-link results. Around the same time, Google’s Knowledge Graph and “People Also Ask” boxes also emerged, offering factual summaries and related Q&A pairs directly on the search page. These features signaled a shift: Google was no longer just a gateway to information, but increasingly an answer engine delivering solutions immediately within the search interface. This shift gave rise to the concept of Answer Engine Optimization (AEO). First coined by forward-looking SEOs in the mid-2010s, AEO refers to tailoring your content to be selected as a direct answer in search results (featured snippets, answer boxes, voice assistant replies, etc.). Unlike traditional SEO which prioritizes earning clicks to your site, AEO is about visibility even without clicks – making your content the authoritative answer that Google (or Siri, Alexa, etc.) provides. Key AEO tactics included: Structuring content around questions and answers: Content creators learned to incorporate common user questions as headings, immediately followed by succinct, factual answers. For instance, an FAQ page might ask “How does X product work?” with a 2-3 sentence answer directly below. This format increases the chance that Google will excerpt that answer in a snippet or voice response. Using lists, tables, and steps: If a query is looking for a list (e.g. “top 5 CRM software features”) or a how-to procedure, formatting the answer as a bullet list or step-by-step numbered list improves snippet eligibility.

Early research showed that Google often pulled numbered lists for “How to…” queries and tables for data-driven queries. Schema markup and structured data: Webmasters began adding structured data (using Schema.org tags like FAQPage, HowTo, Recipe, etc.) to make the page’s Q&A content machine-readable. This helps search engines identify and trust the content format, increasing the likelihood of it being featured. For example, marking up an FAQ section with <FAQPage> schema can directly feed Google’s Q&A features. Voice search optimization: As voice assistants became popular, AEO overlapped with voice SEO. The answers needed to be conversational and concise because devices like Google Home would read them aloud. This reinforced writing in a natural, easy-to-understand tone that still contained the key answer in the first sentence or two. People-first, authoritative content: AEO still required quality. Google favored sources that demonstrated expertise and authority (what later would be codified as E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness). So, while formatting was important, content creators also focused on accuracy, citing reputable facts, and providing genuinely helpful answers to earn Google’s confidence. Together, these practices formed the foundation of AEO. Marketers recognized that search was evolving “from a gateway to a destination” – often answering users’ needs without a click. For example, on mobile or voice, a user might get the full answer read out, never visiting the site. The upside was that if your content was the one featured, your brand gained high visibility and implied endorsement by Google. The downside was fewer clicks and less control over how your content was presented. By the late 2010s, the question for SEO professionals had expanded from “How do I rank #1?” to “How do I become the source that Google’s AI or answer box cites?”.

In essence, AEO was an early playbook for what has now become Generative Engine Optimization (GEO) – optimizing content for AI-generated answers. It’s worth noting that many industries started leveraging AEO techniques. E-commerce sites began structuring product pages to answer common questions (return policies, materials, sizing) directly on-page, hoping to capture “People Also Ask” spots or voice answers about their products. B2B and SaaS companies invested in rich knowledge bases and blog content targeting long-tail questions (e.g. “How to improve team productivity in agile software development”) so that their expertise could appear in featured snippets. In doing so, they not only aimed to drive traffic but also to build brand authority by being the answer users hear or see. This was critical in fields where trust and thought leadership drive sales: if a prospective client keeps seeing a SaaS company’s name popping up in answer boxes about, say, data security best practices, it subtly positions that company as an authority before the user even visits their site. Crucially, AEO set the stage for today’s AI-driven search. It trained a generation of content creators to think in terms of answers, not just keywords. The lessons learned – about concise phrasing, structured information, and anticipating user questions – are directly applicable to optimizing for AI summaries and chatbots. As we’ll see, Google’s latest evolution, the Search Generative Experience (SGE), can be viewed as the next logical step in this “answer-first” evolution of search. SGE uses advanced AI to synthesize answers from across the web, but it still relies on well-structured, authoritative source content – exactly the kind of content that AEO practitioners excel at producing.

6.2 Google’s Search Generative Experience (SGE)

By late 2022 and early 2023, the search landscape was shaken by the rise of powerful LLM-based tools like OpenAI’s ChatGPT. Users flocked to these AI chatbots for quick answers and advice, raising an existential question for Google: Would people bypass traditional search for conversational AI? In response to this competitive threat, Google announced the Search Generative Experience (SGE) in May 2023. SGE represents Google’s bold experiment to integrate generative AI directly into search results, providing AI-crafted answers above the list of links. What SGE Looks Like: In the SGE interface, queries that trigger the AI will display an “AI overview” at the very top of the results page, often with a subtle colored background to distinguish it. This overview is essentially an AI-generated summary of the user's query, synthesizing information from multiple web sources into a few paragraphs of conversational text ( ^[1] ) ( ^[2] ). For example, a user searches “best family SUV for safety and fuel efficiency.” In the SGE-enabled Google, an AI snapshot might appear first, saying something like: “For a family-focused SUV that excels in safety and fuel economy, consider models like the Toyota Highlander or Volvo XC90. The Highlander earns top safety ratings (5-star NCAP) and gets about 24 MPG, while the Volvo offers advanced driver-assistance features and ~27 MPG in its hybrid version…” – and so on, perhaps two or three key points per suggested model. This AI answer is clearly labeled as experimental and is boxed off from the standard results ( ^[1] ). Importantly, SGE doesn’t just present AI text alone – it also cites sources and provides links for deeper exploration. Beneath or beside the AI-written sentences, you will see small citation numbers or clickable cards referencing the websites from which the information was drawn ( ^[3] ) ( ^[4] ). Google has emphasized that the AI overview is grounded in content from the open web: “Google will cite the websites it used to help generate the answer,” noted one report on SGE ( ^[3] ). If you click on a citation or one of the suggested links, you’ll be taken to that source website. In many cases, the AI overview also displays a few “featured sources” as clickable panels – these may include an image (from the source page) with a headline or site name. This is akin to an expanded featured snippet, but with multiple sources featured at once ( ^[2] ) ( ^[4] ). Users can interact with SGE in a conversational way. Follow-up queries are supported: underneath the AI snapshot, there’s often a prompt like “Ask a follow-up” or suggested next questions related to your query. If you continue the conversation (for example, following the SUV query with “What about maintenance costs for those models?”), Google will remember the context (the models discussed) and generate a refined answer, possibly citing new sources. Essentially, SGE brings an element of chatbot-style dialogue into search. Google describes it as carrying context from question to question so users can “naturally continue your exploration”. This is a notable shift from the one-and-done query style of classic search. Another notable aspect of SGE is how it handles different query types and verticals . In informational queries (like “how to improve remote team collaboration”), SGE might produce a multi-point summary answer, sometimes even splitting the answer with subheadings or bullet points if relevant. For shopping and product searches , SGE taps into Google’s vast Shopping Graph to give AI-curated product recommendations. For example, a query like “best noise-cancelling headphones under $200” might trigger an AI snapshot listing 2-3 headphone models with key pros/cons and current price ranges, all compiled from product descriptions and reviews on the web. Google revealed that SGE can pull from over 35 billion product listings in its Shopping Graph, which is updated with 1.8 billion changes per hour to keep prices, reviews, and inventory data fresh ( ^[5] ). This means the AI overview for shopping queries can include very up-to-date information – a critical factor for e-commerce. (In our headphones example, the AI answer might say " Model X from Bose – currently around $180 – offers top-tier noise cancellation and 20-hour battery life, according to 1,200+ reviews ," with the text “according to 1,200+ reviews” linked to a source or Google’s own shopping data.) Google has so far limited SGE to certain query categories. It tends to appear for more complex informational searches, comparative questions, broad shopping queries, and advice-like questions – where a synthesized answer is helpful. On the other hand, SGE is less likely to trigger for very simple factual queries (where a standard featured snippet suffices) or highly sensitive queries (medical, financial, or other YMYL topics), presumably because the risk of AI inaccuracies (hallucinations) in those areas is higher. For example, searching “symptoms of diabetes” might still show the traditional snippet or a panel from a trusted health site, rather than an AI-generated overview, due to Google’s caution around medical advice. Indeed, early testing of SGE showed that Google was imposing higher standards for AI answers on YMYL (Your Money or Your Life) topics, and sometimes SGE would simply not appear at all for such queries. As SGE has evolved (still officially in “Labs”), Google has iterated on when to show the AI. By late 2023, reports noted that Google dialed back SGE’s prevalence: one analysis found the fraction of searches with an SGE result dropped from 75% down to about 35%, meaning Google was intentionally not showing AI on many queries unless it was confident in the value added. This reflects Google’s cautious approach – they are gradually adjusting the presence of AI to balance user experience, accuracy, and revenue considerations (more on that shortly). Rollout and Current Status: Initially, SGE was (and still is) an opt-in experiment. Users had to join the Search Labs program to test SGE. After Google I/O 2023 unveiled SGE, a waitlist was opened for US users; by mid-2023 many users gained access. In late 2023, Google expanded SGE Labs access to more regions – by early 2024 it was available in 120+ countries and 7 languages for testers, although notably some markets (like the EU and UK) had limited access due to regulatory considerations. Google planned to run the experiment through 2023, but in a January 2024 update they announced SGE would remain in the Labs testing phase longer than expected . In fact, Google indicated SGE would continue as an opt-in experiment “for the foreseeable future,” rather than immediately rolling out to all users. This decision came after mixed feedback from users and the SEO community about SGE’s readiness. Several factors likely influenced the extended test period. Quality and accuracy issues were a concern – early on, SGE sometimes made factual errors or drew from less-than-authoritative sources, which could undermine user trust. (A Washington Post tech columnist bluntly wrote in mid-2023, “I tried the new Google... its answers are worse,” noting instances where SGE misinterpreted questions and cited low-quality sources or non-sequiturs.) Google, wary of damaging its search reputation, has been fine-tuning SGE’s AI models and citation mechanisms to improve answer quality. They even publicly stated that they hold SGE to a higher quality bar and have extra guardrails for sensitive topics. User experience and performance are another factor. Early testers observed that SGE could be slow to load – sometimes taking several seconds to generate the AI snapshot, which is far from the instantaneous response people expect from Google. Google’s own guidelines aim for <3 second page loads, and SGE in its initial form often overshot this, risking user frustration. The layout was also very dense, pushing traditional search results far down (especially on mobile screens). By late 2023, Google started tweaking the design: for some queries, instead of auto-generating the AI answer, they showed a “Generate AI Response” button that the user could click if they wanted the AI summary. This opt-in trigger on a per-query basis can preserve screen space and loading time for users who don’t need the AI help. It suggests that Google is considering a more conservative integration – perhaps only surfacing the AI when it clearly adds value, rather than for every possible query. Google also had to consider the operational costs and business impact of SGE. Generating AI answers on the fly is computationally expensive (LLMs require a lot of processing power per query compared to a normal keyword search). If SGE doesn’t significantly improve user satisfaction or defend market share, rolling it out broadly could be a costly gamble with little return. Moreover, SGE threatened to disrupt Google’s ad model: initial versions of SGE pushed organic results (and therefore ads) further down, potentially reducing ad visibility and clicks. As one marketing agency noted, the first iteration of SGE had minimal ad integration, sparking concern that Google’s revenue could dip if people got answers without scrolling. Recognizing this, Google soon added ads into the SGE interface , ensuring sponsored links still appear in dedicated slots even with an AI snapshot present. During the 2023 I/O announcement, Google reassured advertisers that ads would remain a “distinct, identifiable part” of the search experience, even as AI is introduced ( ^[6] ). By early 2024, testers observed that some SGE results included ads at the top or bottom of the AI snapshot, labeled as such. Google is clearly treading carefully to maintain its core business while innovating on the UX. So, as of 2024, SGE remains in an experimental phase , not yet the default Google experience for all. Industry experts are split on whether SGE will graduate from Labs into mainstream search soon or whether Google will keep it semi-optional until the kinks are fully ironed out. One thing is certain: Google’s competitors are not standing still. Microsoft’s Bing integrated GPT-4 into its search early in 2023 (see Chapter 7), gaining a surge of interest. Other platforms like Perplexity.ai launched with AI search that cites sources. Even social platforms (TikTok, Reddit) siphon search share for certain demographics, and Amazon remains the go-to for product searches. SGE is widely seen as a defensive move to protect Google’s dominance . In Google’s ideal scenario, SGE will keep users on Google by offering the convenience of an AI assistant within the familiar search page, rather than losing those users to external chatbots or apps. Early results are mixed: some users love the convenience of AI summaries on Google (“it saved me from clicking 5 different links to piece together an answer”), while others find it too authoritative , potentially limiting them from discovering diverse viewpoints. Google’s own surveys found that users appreciated citations and wanted even more transparency into how AI answers are composed – feedback Google has said it is incorporating ( ^[7] ). SGE is also teaching Google a great deal about how AI can complement traditional search. Even if the exact current form of SGE doesn’t last, the core idea of AI-assisted search is here to stay. And for marketers and content creators, SGE offers a preview of a new kind of search results page – one where getting your content referenced in the AI overview could become as coveted as a Page 1 ranking. In the next sections, we’ll examine Google’s AI model strategy (Bard and Gemini) and then dive into how to optimize for these generative results.

6.3 Google’s Bard Chatbot and the Gemini AI Model

While SGE transforms the Google search interface, Google has a parallel AI initiative in its standalone chatbot Bard . Bard was Google’s answer to ChatGPT – literally. When ChatGPT’s popularity exploded after late 2022, Google fast-tracked Bard’s development and launched it to the public in March 2023 (initially to users in the US and UK). Bard started as a conversational AI service, separate from Google Search, accessible at bard.google.com. It was powered originally by LaMDA , a Google-developed conversational language model. Early reviews of Bard in spring 2023 noted that it was less impressive than ChatGPT-4 in some areas, often providing shorter and sometimes less accurate answers. Google quickly iterated: by May 2023, they announced Bard would be powered by a more advanced model, PaLM 2 , improving its capabilities (especially in coding and math). But the real leap was yet to come – with Gemini . Gemini is Google’s next-generation foundation model, developed by the Google DeepMind team (a collaboration after Google merged DeepMind with its Brain team). In December 2023, Google officially introduced Gemini 1.0, heralding it as “our largest and most capable AI model” to date. Gemini is designed to be a direct competitor (or successor) to OpenAI’s GPT-4. It’s notable for being multimodal from the ground up , meaning it can process and generate not just text, but also images, and even understand modalities like audio and video. This multimodality is an advantage – for instance, Gemini can analyze an image you upload and have a dialogue about it (describe it, answer questions about objects in it, etc.), a capability GPT-4 introduced with its Vision update but that Google aims to excel in natively. Google has built Gemini in different size tiers to serve various use cases: Gemini Ultra (the largest, most powerful model intended for highly complex tasks), Gemini Pro (a mid-tier model optimized for a wide range of tasks efficiently), and Gemini Nano (a smaller version that can even run on mobile devices for on-device AI). This tiered approach is similar to how OpenAI has GPT-4 for premium use and GPT-3.5 for lightweight tasks, but Google is explicitly optimizing Gemini variants for everything from cloud data centers down to smartphones. For marketers, this hints at a future where advanced AI might be embedded directly in devices and apps (imagine AR glasses doing instant visual search with Gemini Nano, for example). Crucially, by early 2024 Google began integrating Gemini into Bard. In fact, Google started to brand Bard’s updates under the Gemini name – so much so that some tech observers refer to the latest version of Bard as effectively “Gemini Chat” . One tech site noted: “Bard AI, Google’s chatbot, has since been renamed Gemini, unifying its generative AI technologies under the Gemini name” . In practical terms, Google launched “Gemini Pro” as the model behind the free Bard and a more powerful “Gemini Ultra” behind a premium version called Bard Advanced (which, similar to ChatGPT Plus, comes with a subscription fee). By March 2024, testers reported that the paid Gemini Ultra (Bard Advanced) could go toe-to-toe with GPT-4 in quality: “Gemini Ultra … provided marginally better responses than GPT-4… as well as better imagery” in one head-to-head test. The free Gemini (Pro) was also found much stronger than the free ChatGPT (GPT-3.5). So what sets Bard/Gemini apart from ChatGPT in the search context? A few points: Real-time information access: Perhaps Bard’s biggest differentiator is its deep integration with Google’s search index. Bard has the built-in ability to “Google” things as needed. If you ask Bard about a very recent event or something that’s not in its training data, it can fetch up-to-date information from the web. (ChatGPT’s base model, by contrast, has a knowledge cutoff – as of 2023, GPT-4’s training data goes up to September 2021, and it requires plugins or a separate browsing mode to get newer info.) For example, ask ChatGPT free version “Who won the 2024 World Cup?” and it cannot answer from its base knowledge (since that event is beyond its training); ask Bard, and it will search Google and tell you the result, citing a news source. This ability is thanks to Bard’s integration with Google Search – essentially a form of Retrieval Augmentation. Gemini was designed with real-time retrieval in mind , and indeed Google positions it as being able to access “the internet” as needed. The result for marketers is that Bard/Gemini may provide more current answers (stock prices, today’s weather, latest product releases, etc.) and thus content that is kept current has a better chance of being surfaced by Bard. Integration with Google’s ecosystem: Beyond just web search, Bard has been woven into other Google services. In late 2023, Google introduced Bard Extensions , allowing the chatbot to pull information from a user’s Google Workspace apps – like Gmail, Google Drive, Docs, Maps, etc. ( ^[8] ). For instance, Bard can summarize the content of your Google Docs, or draft an email using info from your Gmail, if you grant it permission. ChatGPT offers plugins that can do some similar things (and Microsoft integrates GPT-4 into Office 365 in some ways), but Google leveraging its own ecosystem is powerful. For marketers, this means Bard could become a kind of personal assistant that straddles both personal data and web data. Imagine a scenario: a user planning a business trip asks Bard (Gemini) for “Recommend me some project management SaaS tools I can try during my flight to improve team workflow.” Bard might not only list some tools (drawing from web content) but also note “I see in your Google Drive that your team uses Trello currently – one alternative to consider is Asana, which offers XYZ features…” etc. This level of personalization is on the horizon and could transform how product discovery happens via AI. Multimodal capabilities: As mentioned, Gemini is multimodal. Already, Bard can accept image inputs – for example, you can show Bard a photo of a chart or a math problem and ask for analysis. And Bard can generate images via integration with Google’s Imagen model or third-party generators (Google announced in 2024 that Bard can create images through a partnership with Adobe Firefly). This out-of-the-box image generation is not something ChatGPT offered natively (OpenAI relies on DALL-E plugins). So, Bard is positioning itself as a more visual assistant. A practical use-case: an e-commerce marketer could ask Bard, “Generate an image of a modern office desk setup featuring our product” for a mockup – Bard could do it. Or a user could upload a screenshot of an error message and ask Bard for help, and Bard (via Gemini) could read the image text and provide guidance. These features make Bard more versatile and potentially more engaging for users, keeping them in Google’s orbit. Factuality and accuracy improvements: Google has touted Gemini’s performance on knowledge benchmarks. For example, Gemini Ultra scored 90% on the MMLU academic benchmark, making it the first model to exceed human expert performance on that test. This benchmark covers a wide range of subjects (math, history, law, etc.), indicating Gemini’s breadth of knowledge. Additionally, Google claims Gemini has advanced reasoning capabilities – it can “think through” problems better, rather than just blurting out the first answer. In practical terms, one early comparison found “Gemini’s answers often provided more nuance and context than ChatGPT’s” , albeit sometimes at the cost of being a bit wordier. Google is clearly aiming for an AI that is both knowledgeable and context-aware . If Gemini-powered Bard can reduce hallucinations and cite sources more reliably, it might alleviate one of the big concerns users and publishers have with AI answers. Competitive positioning: By 2024, ChatGPT had over 1.6 billion monthly visitors (as of Feb 2024) and was a household name ( ^[9] ). Google, despite being a leader in AI research, was seen as playing catch-up in the public eye. The launch of Gemini and its integration into Bard is Google’s attempt to leapfrog. Early community tests (some of which leaked on forums) suggest that Gemini is extremely capable – for instance, one comparison of “Gemini vs GPT-4 Turbo” in late 2023 found that while GPT-4 might still be a bit better in strict accuracy, Gemini was faster and produced more human-like, creative outputs. Google also has the advantage of scale and distribution – the moment they decide to push Bard/Gemini to all Android phones or Chrome browsers, the user base could explode. We may soon see Bard’s capabilities (like a “search with Bard” option) integrated more into core Google products. In the search context, Google could eventually unify the Bard experience with SGE – for example, a user might switch from the AI snapshot to a full “Bard chat” within search to further discuss the query. In fact, some experiments in late 2023 showed a conversational mode directly in the Google app , which is essentially Bard in Search. From a marketing perspective , the rise of Bard and Gemini means that optimizing for Google’s ecosystem isn’t just about classic SEO, but also understanding how your content might be accessed and presented by an AI. For instance, if Bard is summarizing information about your company (say, pulling from your About Us page or recent news articles), you’d want to ensure those sources are accurate and highlight the points you’d want a summary to include. We might soon consider “ LLMO ” (Large Language Model Optimization) as parallel to SEO – making sure our content is digestible and favorable to AI models like Gemini (we’ll discuss optimization in the next section). To give a concrete example, consider a SaaS B2B company offering project management software. In the old world, they’d focus on SEO for keywords like “best project management tool” to rank their blog or comparison page. In the new world with Bard, a user might ask the AI, “Which project management software is best for small teams? Explain why.” Bard will draw on its training data (which includes tons of web content, maybe including tech review sites, user forums, the SaaS company’s own content if it’s prominent, etc.) plus any real-time info (maybe Gartner’s latest report or Reddit discussions). If our SaaS company has done a great job at publishing high-quality, fact-rich content (like a detailed comparison of tools or noteworthy case studies), Bard might incorporate points from it: e.g., “According to a case study by AcmeCorp (a project management SaaS), small teams often struggle with tools that lack integrations. AcmeCorp’s software emphasizes integration with Google Drive and Slack ( ^[10] ) ( ^[11] ).” This would be a huge win – the AI essentially quoting the company’s own content as authoritative. On the flip side, if a competitor’s content or a third-party blog is more visible to the AI, our company could be left out of the narrative, even if they have a great product. This scenario is exactly what some companies have faced: one case study showed a B2B firm was frustrated that competitors with “inferior” products were being summarized and cited by AI for key queries, while they were absent ( ^[10] ). The firm had to proactively adjust their content strategy to get included (we’ll revisit this case study later) ( ^[11] ). In summary, Google’s strategy with Bard and Gemini is to ensure it has the best underlying AI model and a compelling chatbot interface to keep pace with or surpass OpenAI’s offerings. For users, this means more powerful AI capabilities at their fingertips (especially if you’re in Google’s ecosystem). For online businesses and marketers, it means preparing for a world where both search results and chatbot answers might drive discovery of your brand. Content needs to be optimized not just for the ten blue links, but for the AI dialogues and overviews that millions of people will increasingly rely on.

6.4 Optimizing for Google’s AI-Powered Results (SGE and Beyond)

Google has stated that, from a webmaster’s perspective, nothing fundamentally new needs to be done for SGE – in other words, the same SEO best practices that help your content rank well will also help it be included in AI overviews. In late 2023, Google’s Search Liaison team reassured site owners that they weren’t introducing any special meta tags or SGE-specific ranking algorithms; the AI overview draws from Google’s regular index and ranking signals. “If you’re producing helpful, people-first content, you’re already doing the right thing,” is the general message from Google. That said, optimizing for an AI-driven results page does put a new lens on some familiar tactics. Essentially, we need to ensure our content is accessible, understandable, and authoritative to both search engine algorithms and the AI systems generating summaries. Let’s break down the key optimization considerations:

6.4.1 Ensure Crawlability and Access for AI

First and foremost, your content must be indexable by Google. The AI can’t summarize or cite what it can’t read. This might sound obvious, but it’s worth double-checking: are any critical sections of your site blocked by robots.txt or meta noindex tags? Ensure that your robots.txt isn’t disallowing important directories (like /blog/ or /knowledge-base/). Likewise, avoid using heavy client-side rendering for content that could be an answer – if the content only loads via JavaScript after page load, Google’s crawler might not always see it. A pure HTML text version (or server-side rendering) of key content is safer for ensuring the AI has access to it. It’s also important to note that currently there is no way to opt out of being included in AI overviews without opting out of search altogether. Some publishers, concerned about content being used by AI without clicks, asked Google for a “SGE opt-out” meta tag. Google’s answer was that SGE is just a new way of presenting search results, so the usual rules apply – if your page is indexed in search, it may be used in an AI answer. The only surefire way to exclude content from AI summaries would be to add a noindex (which would remove it from Google search entirely) – obviously not a desirable solution for most. There has been discussion of a future Google-Extended tag to control use of content in AI training, but for search outputs like SGE, nothing granular exists yet. Therefore, for now, assume all your indexed content is fair game for AI summaries. This is actually an opportunity: it means even pages that might not rank #1 on their own could still get visibility if they contain a relevant snippet that the AI finds useful.

One more technical note: make sure your content delivery is fast. SGE currently has some performance issues, and Google will likely favor content that can be retrieved and parsed quickly (especially if it’s assembling answers on the fly). So, core web vitals (good LCP, TTFB, etc.) and overall site speed indirectly help. If your page is very slow or unresponsive to Googlebot, the AI might skip it in favor of a faster source when constructing an answer, especially for time-sensitive queries.

6.4.2 Use Structured Data and Semantic HTML

Structuring your content clearly – both for human readers and machine parsing – is vital for AI optimization. Structured data markup (Schema.org) can give Google explicit clues about the content on a page. For instance: Adding FAQPage schema to a list of Q&As on your site could increase the chance that one of those Q&As is pulled into an AI answer (or at least appears in the “People Also Ask” which the AI might consider). Google’s documentation notes that using schema for FAQs, how-tos, products, etc. helps make your information eligible for rich results, and by extension, these structured pieces are easier for an LLM to digest and trust. HowTo schema can highlight step-by-step instructions. An AI overview may not list all your steps, but it might say “There are 5 key steps to do X” and cite the source – which could be you, if your how-to is marked up and indexed. Product and Review schema are crucial for e-commerce content. As mentioned, SGE’s shopping answers pull in product specs, prices, and ratings. That data often comes from Google’s Shopping Graph, but Google also scrapes schema on product pages for things like aggregate ratings and price ranges. Ensuring your product pages have up-to-date Product, Offer, and Review structured data means the AI is more likely to use your page as a source for product recommendations or comparisons. Article or BlogPosting schema can help with news or informational content. While not directly guaranteeing inclusion, it provides clear metadata (like author, publish date) that could lend credibility. Google’s AI might be programmed to prefer up-to-date info, so seeing a recent date on an article (via structured data) could help, for example, an AI answer about “2024 economic outlook” to choose your article if it knows it’s fresh.

Beyond formal structured data, pay attention to your HTML structure. Use descriptive headings (<h1>…<h2>…) for sections, especially for any question-answer pairs. If you have content that answers a specific question, make that question an <h2> and the answer a paragraph below it. Not only does this align with AEO tactics (featured snippet optimization), it also makes it easy for an LLM to extract the Q&A pair. Consider a page structure optimized for AI: an initial concise answer or summary (that could be read on its own), followed by details. In fact, some SEO experts suggest using an “inverted pyramid” style of writing for GEO – put the direct answer or conclusion at the top (which might be what the AI grabs), and then elaboration afterwards. This way, if the AI overview quotes your content, it’s quoting the punchline you want, not an incomplete mid-explanation sentence. Lists and tables: If applicable, use HTML lists (<ul>, <ol>) for lists of recommendations, features, pros/cons, etc. SGE has been observed to present a lot of information in bullet form (the AI itself often formats key points as bullets or numbered steps). If your content already has a well-structured list, the AI might incorporate it or at least more easily parse it. For example, a site listing “10 benefits of using CRM software” in a neat <ol> could be directly distilled by the AI into a summary, citing that site. Similarly, tables (for structured data comparisons) can be read by AI – if you have a comparison table of specs, the AI might use it to call out a specific number (“the Nikon camera has a 24.2 MP sensor vs 20 MP on the Canon” and cite the source).

To give a real-world example: a travel blog might have a well-formatted comparison of Bryce Canyon vs Arches National Park for families (recall the SGE example query from Google’s demo). If that blog uses clear subheadings like “## Bryce Canyon for Families” and “## Arches National Park for Families,” and maybe a bullet list of pros for each, then SGE’s AI might pull points from each section to form its answer. In Google’s own demo, the AI said, “Bryce Canyon has more toddler-friendly short hikes, while Arches requires more scrambling; Bryce’s elevations mean cooler temps, etc.” – these details had to come from someone’s content. If your blog had a section stating “Bryce Canyon: Several short, easy trails like Mossy Cave are suitable for toddlers.” and another stating “Arches: Some trails involve rock scrambling not ideal for kids under 5.”, you greatly increase your chance of being one of the cited sources in the AI snapshot. And notably, those points need to be explicit and crisp – an AI might not infer or combine scattered info as reliably as we think. It’s often doing pattern matching and summarizing in a local way. So make important facts stand out (bold them, list them, or at least not bury them in fluff).

6.4.3 Focus on Helpful, Authoritative Content (E-E-A-T)

Google’s core advice with SGE is that “helpful, people-first content” will be rewarded. This aligns with the Helpful Content system and the E-E-A-T principles (Experience, Expertise, Authoritativeness, Trustworthiness) from its search quality guidelines. Why does this matter for AI results? Because Google’s generative AI is likely using signals of authority and accuracy when choosing what to include in an overview. In fact, Google has hinted that their SGE algorithms consider page quality and ranking just as a normal result would. One agency noted: “Currently, the best way to appear as one of these relevant [AI overview] pages is to be one of the top-ranked results”. In other words, if your page already ranks on page 1 for the query (or a closely related query), it’s a prime candidate to be cited. The AI isn’t intentionally seeking obscure sources; it’s drawing from the cream of search index for that topic (with perhaps a few exceptions as noted earlier). Thus, the traditional pillars of SEO remain crucial: good content that satisfies user intent, backed by backlinks/authority, and demonstrating expertise. Ensure your content is accurate and well-researched. AI has a habit of confidently spouting wrong info if its sources are poor. Google will try to avoid citing sources that have inaccuracies or thin content because that reflects back on the AI’s quality. If your site has reviewed by experts or author bios with credentials, that might indirectly help (for example, a medical article written by a doctor might be deemed more trustworthy by Google’s algorithms and chosen for an AI summary on a health question). Additionally, consider adding references and data in your content. If you provide a statistic or important claim, cite a source (and link to it). This might sound counter-intuitive (why send people elsewhere?), but it can boost your credibility.

There’s a scenario where the AI might even mention that data point and your site together. For instance, “According to YourSite, 65% of small businesses saw ROI from CRM within 6 months.” If that statistic is real and you cited a study, the AI might choose to use it and cite you (and possibly the original study if it knows it). We are essentially treating the AI as another consumer of our content – one that has read millions of pages. If your page consolidates useful facts or insights in one place, the AI may prefer it to having to pull from multiple different pages. Original research and unique insights are also key. In a world where AI can generate generic content easily, having something truly original (a proprietary survey, a case study, a unique viewpoint) sets your content apart. Google’s algorithms (and by extension SGE) aim to highlight “information that AI cannot easily produce itself”. Long-term, content that just rehashes common knowledge may be deemphasized, while content with genuine expertise or firsthand experience will be valued. So, for example, an e-commerce site could publish internal data about how customer satisfaction improved after using their product, or a SaaS company might release survey results from their user base – these kinds of things, if cited by others, increase your authority. They also give the AI something novel to latch onto (e.g., “Brand X’s report says 3 in 4 remote teams work across time zones, highlighting need for scheduling tools…” – that could feed an AI answer on remote work challenges). In practice, this means content quality and depth should not be sacrificed. Just because an AI summary will be brief doesn’t mean you only write brief content.

It means the most important parts of your content should be easily extractable, but you still want to have depth for the users who do click through (and for overall topical authority). Topical authority is another concept: cover your subject comprehensively across your site. If you run a fintech blog, having a cluster of well-written articles on related subtopics (and interlinking them) helps Google view you as authoritative in that domain. We suspect that SGE’s selection of sources leans towards sites that have authority on the topic at hand (e.g., it might pull from a specialist site on a medical query rather than a generic site). In one observation, early AI overviews about programming questions often cited sites like Stack Overflow or official docs (authoritative sources) rather than random blogs. So, building your site’s topical authority and reputation (through content and links) is key to being one of the chosen sources in AI results.

6.4.4 Optimize for Conversational Queries and Prompts

In the age of generative AI search, users are likely to phrase queries more conversationally or specifically . Instead of the old terse keyword searches, people might ask the AI, “What should I do if my e-commerce site’s traffic drops after SGE rolls out?” – something they might not type in the regular Google. Marketers should anticipate these natural language queries and embed answers to them in their content. This is akin to “Prompt SEO” – optimizing content to align with the kinds of questions users will pose to AI assistants ( ^[12] ) ( ^[13] ). A practical tip: maintain an FAQ section or Q&A content that covers the “who, what, why, how, when” around your niche. If you have a comprehensive FAQ, each question there might match a conversational query a user asks the AI. For example, if you sell eco-friendly home products, an FAQ question like “How can I reduce plastic waste in my kitchen?” with a well-crafted answer could be exactly what someone asks Bard or SGE. Your content might then be used in the AI’s answer (with a citation). It can also be useful to mirror likely user phrasing . If people often type questions like they’re talking (e.g., “What’s the best way to learn SEO in 2025?”), consider having a blog post titled “The Best Way to Learn SEO in 2025 – Answered” or literally phrasing a header as a question: “What Is the Best Way to Learn SEO in 2025?”. This gives the AI a clear signal that your section answers that exact query. It’s similar to old featured snippet targeting, but even more conversational. Tools that analyze People Also Ask questions and community forums (like Reddit, Quora) can give insight into how real users ask things – use those insights in your content creation. Also, consider that AI chat follow-ups might combine concepts. For instance, a user might start general (“give me tips to improve my website’s UX”), then follow up specifically (“how about for mobile users?”). If your content covers both general and specific subtopics (desktop vs mobile UX), the AI could draw from different parts of your site across turns. Ensure your site’s internal linking and structure connect these related topics so Google knows you cover them all. A pillar page with links to specialized subpages can work well. In some cases, embedding likely prompts verbatim in your content can be useful. This is a bit experimental, but some have tried adding, say, a hidden HTML comment or very small footer text like “ Prompt: What are the benefits of using [Your Product]? ” followed by a brief answer. The idea is to literally feed the LLM an exact Q&A. Whether this works or is sustainable is unclear (and doing it at scale could be seen as manipulative if overdone). A safer approach is just making sure your content naturally includes the question phrasing in visible text.

6.4.5 Technical SEO and Site Signals in the Generative Era

Beyond content and schema, technical SEO factors still matter in an AI context. Google has indicated that core ranking signals apply to AI selections, which likely includes things like PageRank (backlinks), content relevance, freshness, and even user experience signals. So continue to invest in: High-quality backlinks from reputable sites. If many sources link to your content as a reference, Google’s algorithms are more likely to deem it trustworthy and therefore safe to include in an AI answer. Off-page authority (see Chapter 11 on digital PR and authority signals) can indirectly boost your presence in AI overviews. Freshness: For topics where information changes over time (tech, finance, trends), updating your content regularly can be crucial. Google’s AI will incorporate real-time info for sure via search, but if your page is outdated, it may favor citing a fresher source. We’ve seen AI overviews include phrasing like “As of October 2024, ...” which implies it looks for timestamped info. So keep dates updated and content current. If you have evergreen content, consider adding a “Last updated” date in the text if you refresh it – the AI might pick that up. Robust website structure: Ensure your site is well-organized so that Google can understand context. Use clear URL structures, breadcrumbs, and category pages that demonstrate how topics are related. A well-structured site helps Google’s index and also means if the AI is looking for related info (to answer a follow-up question for example), it might find it on your site (since your related content is interlinked and easy to crawl). Preventing errors: Fix or redirect broken links, resolve 404s, etc. If an AI tries to cite your page and it’s not reachable, that’s a missed opportunity. Google’s index might drop it if it’s consistently erroring.

Also, if you move content, use redirects so that any accumulated signals (including being a known source for an answer) carry over to the new URL. Finally, monitor what queries and answers your site is appearing for. Google Search Console in 2023 added some reporting for “AI Features” impressions (for sites in SGE) – it would show if your site was cited in an AI overview and for what query (though in limited fashion). These tools will likely improve. By keeping an eye on this, you can learn which content of yours is resonating with the AI and expand on those areas or refine them. Conversely, if you notice competitors appearing in AI answers where you think your content is better, analyze why – do they have better structured data? More concise answers? Perhaps their page is just slightly more on-target for the query. This insight can inform your content optimization. In summary, optimizing for Google’s AI results is an evolution, not a revolution, of SEO practices. It demands a renewed focus on clarity, structure, and truly useful content. The upside is that many of these practices (like AEO) were already being adopted by savvy SEO professionals. The key difference now is thinking about how an AI bot reads and synthesizes your content, in addition to how a human or a traditional crawler would. By ensuring your site is technically sound, your content is well-structured and authoritative, and by anticipating the kinds of questions users (or the AI on their behalf) will ask, you position yourself to be a prominent player in the generative search landscape.

6.5 Opportunities and Threats of Generative Search for Marketers

The emergence of AI-generated search results brings a mix of exciting opportunities and significant challenges for online marketers. It’s a classic double-edged sword scenario: on one hand, generative search can dramatically expand a brand’s visibility by pulling in information from all corners of the web (potentially giving smaller or niche sites a chance to shine); on the other hand, it threatens the traditional traffic model by often satisfying users’ queries without a click , which could erode website visits and the ability to convert or monetize those users. Let’s break down the key opportunities and threats, with real-world context:

6.5.1 The Zero-Click Paradigm – Threat to Traffic, Challenge to Adapt

One of the most obvious impacts of AI answers is the reduction in click-through rates (CTR) for organic results. If an SGE overview or Bard gives the user exactly what they need, they may have little incentive to click any of the source links. This trend was already underway with featured snippets and Google’s direct answers (zero-click searches have been rising for years), but AI takes it to another level by handling more complex queries end-to-end. Early studies have tried to quantify this potential traffic loss. A notable analysis by Search Engine Land in mid-2023 modeled the impact of SGE across 23 different websites. The findings projected an organic traffic drop between 18% and 64% on average if SGE-like AI results rolled out widely. Some websites in the study could lose as much as 95% of their organic traffic in the worst-case scenario, essentially decimated if their content is fully cannibalized by AI answers. That said, a few sites were projected to gain traffic (up to +219%) because the AI might highlight them more than traditional search did. This wide range indicates that the impact is not uniform – it heavily depends on the type of content and queries a site targets. Informational, how-to, and FAQ-style content is most at risk for zero-click consumption (since AI can often satisfy those queries). For example, a simple recipe site could see fewer visits if SGE lists the key ingredients and steps right on the results page. Transactional or interactive content (like tools, calculators, product purchases) may be less impacted since users still have to click to complete an action. Marketers should brace for potentially lower raw traffic from SEO, especially for top-of-funnel informational queries. However, traffic quality might shift more than quantity.

Consider that users who do click through after seeing an AI summary might be deeper in the funnel – because either the AI spurred their curiosity for details or the query was such that the AI encouraged them to “dig deeper” on a specific link. In other words, if the AI overview didn’t fully answer their need, the click they make could be more qualified (they’re specifically interested in what your site can offer beyond the summary). Some optimists suggest that while overall impressions and clicks might drop, the conversion rate of the remaining clicks could rise (since casual info-seekers might have been satisfied by the AI, leaving the more motivated users to click). It will be crucial to monitor metrics like engagement, conversion, and user behavior for AI-referred traffic versus traditional traffic. Nonetheless, the threat is real for many content businesses. Publishers that rely on page views for ad revenue are particularly worried. If a site’s painstakingly written article on “10 ways to reduce mortgage payments” is distilled by Google’s AI into a neat bullet list that requires no click, the site loses the ad impressions and possibly the affiliate link clicks it might have gotten. News and how-to publishers are already strategizing about this – some are exploring alternate revenue models or focusing more on content that AI can’t easily display (like interactive tools, videos, or very in-depth pieces).

6.5.2 Brand Visibility and Authority – Opportunity in the AI Spotlight

On the flip side, being featured in an AI overview can significantly boost brand visibility and credibility , even absent a click. It’s akin to being quoted as an expert in a news article – even if readers don’t immediately run to your website, your name being present lends you authority. In fact, one marketing commentary likened AI Overviews to word-of-mouth recommendations : “Google AI Overviews are like friends who save you time by giving quick info… For brands, AI Overviews can increase your brand visibility in the same way positive reviews boost your reputation. Brand visibility builds recognition and trust – pillars that turn strangers into customers.” ( ^[14] ) ( ^[15] ). In other words, if your brand is consistently popping up as a cited source for answers in your domain, users begin to trust and recognize you, even if they haven’t visited your site yet. Imagine someone sees an AI answer about fitness that ends with “– info provided by MyFitnessSite.com.” The user might not click, but subconsciously they register MyFitnessSite.com as a knowledgeable source. After a few times, they might directly seek out that site or feel more comfortable converting when they do land there, because the AI (an ostensibly neutral party) has effectively endorsed it by citation. In marketing, this is often called “brand imprinting” – repeated exposure through trusted channels builds familiarity. There’s also a long-term SEO angle: if AI overviews drive what some are calling “ implicit traffic ” (exposure without click), it could still lead to branded searches later. A business might see an increase in people searching their brand name or typing their URL directly after frequently being mentioned in AI results. Those are highly valuable because they often convert better. Another opportunity is that AI can surface a wider range of sources than traditional top-10 results . We touched on this: SGE might pull relevant bits from sites that weren’t ranking #1. Maybe they were ranking #5 or #10 or beyond, but had a perfect sentence the AI needed. For niche sites or new entrants, this is a chance to get in front of users without beating giants on every keyword. A 2024 report noted that “marketers have noticed not all recommendations [in AI overviews] are from the top search results” , indicating Google sometimes includes diverse sources beyond the usual suspects ( ^[7] ) ( ^[16] ). For example, an AI answer about a programming question might cite a specific developer’s blog that had exactly the solution code, even if that blog isn’t StackOverflow or doesn’t have massive PageRank. In normal search, that blog might be buried, but the AI “fanned out” to find the precise info needed. This leveling of the playing field is an exciting prospect for high-quality content creators who might not have had the SEO clout to appear on page one. It encourages focusing on depth and specificity – if you answer a very specific question really well (even as part of a larger article), you might get picked up by AI for that piece of info. To harness this opportunity, brands should ensure their name and website are clearly associated with their content . For instance, use a consistent brand name in bylines or within content if appropriate (“According to a study by [Brand], ...”). The AI often will cite the source with the site name or sometimes the publication name. You want to be sure it’s attributing it to your brand and not just a generic description. Schema markup can help here too (e.g., publisher name in Article schema).

6.5.3 New Content Strategies – From Click Acquisition to Information Shaping

Generative AI in search will push marketers to adjust their content strategy and KPIs. The goal is no longer just to get the click, but to shape the conversation even if the click doesn’t happen. This means success might be measured in part by being referenced in AI outputs, not just by traditional traffic metrics. We’re moving toward what some call “post-click SEO” or “zero-click marketing” ( ^[17] ) ( ^[18] ). In zero-click marketing, the focus is on ensuring your brand/message reaches the user within the search interface itself. One strategy is to provide content that complements AI rather than competes with it . For instance, AI can summarize general knowledge quickly – but it might encourage users to click through for details, examples, visuals, or interactive elements . If you anticipate that the basic answer to a question will be handled by SGE, consider what extra value a user would get by coming to your site. Then make sure to highlight that on the site so that the AI overview might even hint at it. For example, a cooking site might know that an AI can list ingredients and basic steps for a recipe. So on the recipe page, they include a tip like “watch our 2-minute video for a clever chopping technique” or mention “use our interactive ingredient scaler for different serving sizes”. The AI might say, “For detailed technique, see Video – [Site Name]” if it was trained to note that, or at least the user sees there’s something on that site beyond the text. While currently AI overviews don’t generate new media, in the future they might. But at least for now, unique content like videos, infographics, downloadable templates, quizzes, etc., cannot be fully conveyed in a text summary. Those are click-worthy enticements . Additionally, consider creating more “AI-friendly” content pieces such as: One-stop guides that cover a topic comprehensively. AI prefers not to have to merge too many sources, so if your single page covers multiple facets of a question, it might just use your page alone for an answer, listing you as the sole citation (a big win). We saw AI answers that sometimes only cited one source for a chunk of text. Concise summaries within pages (maybe as a highlighted box or intro paragraph) that the AI could grab. Essentially giving the AI what it needs on a silver platter, and then providing elaboration for the human readers. Conversational tone where appropriate. If the AI is looking for a phrasing to use, content written in a straightforward, conversational manner might be more directly usable. Extremely academic or complex language might get reformulated by the AI (losing your wording) or ignored if it’s not easily digestible. Up-to-date content : We mentioned freshness – but also jumping quickly on emerging questions. When a new trend or problem arises (say a new software update causing issues that people will ask AI about), being among the first to publish a clear answer increases the odds you become the referenced source before others catch up. Essentially, there’s a first-mover advantage in feeding the AI certain answers (until it retrains or shifts to new sources). Let’s not forget local and transactional opportunities too. Generative search will likely extend to local search (“What’s a good pizza place near me that’s open now?” might yield an AI answer listing a couple of restaurants with details). Ensuring your Google Business Profile data is accurate is critical so that if AI pulls operating hours or reviews, it’s correct. If you’re a local business, the AI overview might list you – even if previously you struggled to rank in the pack – based on specific attributes like “kid-friendly” or “pet-friendly” gleaned from reviews or content. This is an opportunity to influence what the AI says by managing your online reputation (encourage happy customers to mention specific positives in reviews, etc., because the AI could summarize common sentiments). E-commerce stands to face threats in terms of affiliate site traffic (as mentioned, aggregator sites might see less clicks if AI lists products). But retailers could gain if the AI funnels users directly to product pages. Google’s SGE, for example, often provides buying options within the overview. If you’re a merchant and your product is one recommended by the AI, you might actually see higher conversion because it’s like a trusted suggestion followed by a direct link to purchase. This raises the stakes for feed optimization and product SEO – you want your products to be well-represented in Google’s Merchant Center, with great reviews, so they get picked by the AI as part of the answer.

6.5.4 Mitigating the Risks – Strategies to Encourage Clicks and Engagement

To address the traffic threat, marketers can deploy several tactics to make their snippet in the AI overview as enticing as possible – essentially to “earn” the click even when the answer is partly given: Tease additional value: If appropriate, ensure the text that might be pulled (often the first 1-2 sentences of your answer) hints at more to be found. For example, “The three main strategies are A, B, and C. Each comes with unique challenges – for instance, Strategy A [something intriguing]...”. The AI might include “Each comes with unique challenges – for instance, Strategy A involves hiring new staff (according to SiteName)…”. A curious user will realize the site likely explains all the challenges, not just Strategy A, and may click to learn them. This is delicate – you must still answer the question, but you can also create a curiosity gap . Provide rich media on click-through: Users who click after an AI answer may be looking for depth or confirmation. Greet them with something that validates their click. If they come from an AI summary about, say, “tips for reducing mortgage payments,” and your site has an interactive calculator or a detailed case study that obviously couldn’t be in the summary, they’ll feel rewarded for clicking. This reduces pogo-sticking and shows Google (and users) that your site delivers beyond the snippet. Optimize titles and meta descriptions (for the links that do show): In SGE, when sources are shown as cards or link bubbles, often the page title or a truncation of it is visible. A compelling title may convince a user to click “Learn more on [Site Name].” If all sources have generic titles and yours is catchy or promises something extra, you could win the click. For example, if the query is answered by AI and sources show up, a title like “Ultimate Guide to X (Free Template Inside)” might draw a click over a title like “Guide to X”. Leverage branding: If you have a strong brand and logo, that could help. In SGE, some source cards show a thumbnail (like a favicon or image from the page). Ensure your favicon is recognizable. Also, if your brand is known for quality, users might click your source over others. This is more about overall brand building – one reason to invest in content marketing, PR, etc., outside of just SEO, is so that when your name does appear as a source, people trust it. Monitor and shape the narrative: If the AI summaries consistently misrepresent something or cite a competitor with possibly outdated info, that’s a signal to produce content clarifying that issue (and perhaps even overtly comparing or debunking if tactful). While you cannot directly control the AI, you can put out content that sets the record straight on topics related to your brand. Over time, as the AI training data updates or the retrieval algorithms improve, your perspective might get picked up. For instance, if an AI answer about your product’s pricing is wrong, ensure your site clearly states the pricing in a prominent way. Google’s AI might then pick the correct info next time (and cite you). There is also a defensive strategy to consider: diversification of traffic sources . If organic search traffic becomes more volatile due to AI, smart marketers will hedge by building up other channels: email lists (so you can reach your audience directly), social media communities, referral partnerships, etc. The js-interactive article noted that “the go-to marketing platform may no longer be exclusively Google. Diversify to other channels… TikTok, LinkedIn, etc., depending on where your audience is” ( ^[19] ) ( ^[20] ). In other words, don’t put all your eggs in the Google basket . While SEO remains crucial, ensuring you have a strong brand presence outside of search will help if AI search changes the rules unexpectedly.

6.5.5 Case Study Reflections: Winners and Losers in Early Generative Search

To ground this in reality, let’s briefly look at a couple of examples from the field: A tech blogging site (we’ll call it TechSite) found in late 2023 that for many “how to fix X error” queries, SGE was providing an answer with a step-by-step from their content – but users hardly clicked through because the overview was sufficient. Their initial traffic from those queries dropped ~20%. However, they noticed an interesting trend: their brand searches increased slightly, and on forums people referenced TechSite’s advice (likely seeing it via SGE). So TechSite pivoted by putting a “video demonstration” on those how-to pages. When SGE began occasionally mentioning “see video at TechSite for demonstration,” their click rate improved. They also doubled down on more complex content (like troubleshooting flowcharts) that the AI could not easily display, ensuring at least some portion of the answer requires a visit. A B2B SaaS company offering data analytics tools saw an initial drop in blog traffic after Bing and Bard started answering questions like “How to build a sales dashboard”. Much of the basic “how to” was given by AI. However, they participated in the Bing Chat and ChatGPT plugin ecosystem to create a ChatGPT plugin for their tool and a Bard Extension . Now, when users ask how to do something, the AI can actually hand off to their product or give a tailored answer using their plugin, often mentioning the brand. This drove qualified leads who directly tried the tool via AI referrals. It’s a reminder that integration with AI platforms (via APIs, plugins) is another opportunity – though outside Google SGE’s domain, it’s part of the wider generative AI landscape. An affiliate content site (monetized by Amazon affiliate links) in the home improvement niche experienced a stark warning: one of their top articles “Best cordless drills 2024” saw traffic sink as SGE’s snapshot literally listed 3 drills with brief bullet points (sourced partly from their content, among others) and users had little reason to click their roundup. In SGE’s cited sources, bigger sites (like Wirecutter) were listed. This affiliate site realized they needed to differentiate. They updated their content to include more in-depth reviews, personal testing insights, and unique photos . They also pivoted to more long-tail content like “Which cordless drill is best for DIY furniture building?” – something AI might not directly answer with a generic snippet. Their strategy is to target queries that require more nuance, and to make their roundup so comprehensive (with pros and cons, user comments, etc.) that a user would want to click to get the full picture. It’s a tough game – some affiliate sites may consolidate or shift focus to either become the authority (so AI picks them) or find topics AI answers can’t fully cover. In conclusion, generative search is reshaping SEO and content marketing , but it’s not an apocalypse for those who adapt. The fundamental need – users seeking trustworthy information and solutions – remains. What’s changing is how the information is delivered and credited . Marketers who focus on earning that credit (being the cited, trusted source) and who think beyond the click (leveraging brand impact and multi-channel strategies) can still thrive. Indeed, those who move quickly to optimize for AI search may capture a competitive advantage, at least in the interim. Generative AI will likely keep evolving – Google Gemini’s improvements might reduce hallucinations, increase the breadth of sources used, or even give more explicit citations (maybe even logos or snippet previews). There’s even talk of AI answers eventually containing embedded content like videos or interactive elements – imagine a future SGE that could play a 5-second clip from a source’s video within the answer. Marketers should stay agile, keep user experience at the center, and treat the AI not as an enemy but as the new intermediary to please. In many ways, we are optimizing for an AI audience on behalf of the human audience. It’s a challenging new frontier, but as we’ve seen from the early cases, those who experiment and learn will find ways to turn these shifts into new forms of success. Sources: Ramp Blog: “What is answer engine optimization?” (July 3, 2025). Google Blog (The Keyword): “Supercharging Search with generative AI” – SGE announcement (May 10, 2023) ( ^[3] ) ( ^[2] ). Search Engine Land: Barry Schwartz, “The new Google Search Generative Experience: Here’s what it looks like” (May 10, 2023) ( ^[1] ) ( ^[5] ). Tinuiti Blog: “Current State of Google’s SGE – What it means for SEO in 2024” (Feb 27, 2024). BlueSoup Agency Blog: “What is SGE and when is it launching?” (Jan 2024). Google Blog: Demis Hassabis, “Introducing Gemini: our largest and most capable AI model” (Dec 6, 2023). Tech.co: “Google Gemini vs ChatGPT 2024: AI Chatbot Head-to-Head Test” (Mar 13, 2024). JS-Interactive: “Will Google AI Overviews Impact Your Brand Visibility?” (Oct 1, 2024) ( ^[15] ) ( ^[7] ) ( ^[16] ) ( ^[19] ). Search Engine Land: Gilad David Maayan, “How Google SGE will impact your traffic – 3 case studies” (Sept 5, 2023). Diggity Marketing: Matt Diggity, “AI SEO Case Study: 2,300% Traffic Increase by cracking AI Overviews” (June 18, 2025) ( ^[21] ) ( ^[22] ) ( ^[10] ) ( ^[11] ).

References

^[1] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[2] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[3] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[4] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[5] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/new-google-search-generative-ai-experience-413533

^[6] Tinuiti.Com Article - Tinuiti.Com URL: https://tinuiti.com/blog/search/search-generative-experience

^[7] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[8] Google Article - Google URL: https://blog.google/products/gemini/google-bard-new-features-update-sept-2023

^[9] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[10] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

^[11] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

^[12] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

^[13] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

^[14] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[15] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[16] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[17] www.yesandbeacon.com - Yesandbeacon.Com URL: https://www.yesandbeacon.com/blog/zero-click-marketing-strategies-search-engagement

^[18] www.rankuno.com - Rankuno.Com URL: https://www.rankuno.com/blog/zero-click-searches-and-their-impact-on-brands-navigating-the-new-seo-landscape

^[19] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[20] Js-Interactive.Com Article - Js-Interactive.Com URL: https://js-interactive.com/sge-impact-brand-visibility

^[21] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

^[22] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

7. Beyond Google – AI Search Alternatives (Bing, Perplexity & More)

The search landscape is no longer dominated solely by Google. A new wave of AI-powered search engines and assistants has emerged, offering users conversational answers, summaries with citations, and innovative features that challenge the traditional “ten blue links” model. This chapter explores the key alternatives beyond Google – from Microsoft’s Bing Chat to independent platforms like Perplexity AI and others – and examines what they mean for marketers in the era of Generative Engine Optimization (GEO). We’ll look at how these AI-driven search tools work, real-world adoption trends in 2024–2025, and practical steps for optimizing content for them. While Google still commands the lion’s share of search traffic, early adopters of AI search are often tech-savvy and influential. Understanding these platforms now can give marketers an edge (and many tactics will carry over to Google’s own AI search features). Chapter 7 Contents: 7.1 Bing Chat’s GPT-4 Integration and Influence – Microsoft’s AI-powered Bing, how it works, and its impact on search behavior and SEO. 7.2 Perplexity AI: The Rise of a Citation-Focused Answer Engine – An overview of Perplexity’s approach to AI search with sources, and its growing user base. 7.3 Other AI-Driven Search Tools (DuckDuckGo, Neeva, You.com, Brave) – Notable alternative search engines incorporating AI, their features, and what trends they set. 7.4 Optimizing for Alternative AI Search Platforms – GEO best practices for these platforms: authoritative content, community presence, and technical considerations. 7.5 Should Marketers Care? – Reach vs. Opportunity – Discussion of the market share and influence of these alternatives, and why investing in them can be worthwhile despite smaller audiences.

7.1 Bing Chat’s GPT-4 Integration and Influence

Microsoft made headlines in early 2023 by integrating OpenAI’s GPT-4 into Bing and launching the new Bing Chat feature ( ^[1] ). This marked one of the first major attempts to embed generative AI directly into a mainstream search engine. Instead of the familiar page of ranked links, Bing’s AI mode delivers a conversational answer that synthesizes information from across the web and presents it in narrative form, with footnote citations linking to sources ( ^[2] ) ( ^[3] ). Users can type natural language questions and receive an AI-generated summary or explanation, often combining information from multiple webpages. Crucially, Bing’s implementation addresses a key concern with generative AI: each answer includes references to the websites used , usually denoted by numbered footnotes that users can click for verification ( ^[2] ). This not only lends transparency and credibility, but also creates a pathway for web traffic – if a user wants more detail or to verify the AI’s answer, they can click the citation and visit the source site. How Bing Chat Works: On the backend, Bing Chat marries Microsoft’s search index (Bing’s web crawling and indexing capabilities) with the language understanding of GPT-4 ( ^[3] ). Essentially, when a user asks a question, Bing’s system retrieves relevant content from live web results, then the GPT-4 model “reads” those results and composes a coherent answer in real time, complete with supporting references. This is a form of Retrieval-Augmented Generation (RAG) in action. Because it uses current web content, Bing Chat can provide up-to-date information , overcoming the “knowledge cutoff” limitation of static trained models. Microsoft initially rolled out Bing Chat via a limited preview (waitlist) in February 2023 and later expanded access, including integration into the Edge browser’s sidebar and Bing’s mobile apps ( ^[4] ). By mid-2023, Bing Chat was accessible to all users of Microsoft Edge, effectively reaching millions by being baked into the default Windows web experience. A New User Experience: The introduction of Bing’s chat mode fundamentally changes how some searches work. Users can now have a multi-turn conversation refining their query – asking follow-up questions or for clarifications – much like they would with a human expert. This conversational capability means search is becoming less of a one-and-done query and more of an interactive dialogue ( “from queries to conversations” , as discussed earlier in Chapter 3). Bing’s AI remembers context within a session, so it can tailor answers based on previous questions, making the experience more intuitive and personalized ( ^[5] ). From the user’s perspective, Bing Chat offers several distinctive benefits: Direct Answers, Fewer Clicks: Instead of scanning multiple sites, the user often gets the answer they need summarized in one go. Bing’s AI-generated response often fulfills the query without the user needing to click any result ( ^[6] ). For example, a query like “What are the health benefits of green tea?” might yield a concise paragraph citing a few authoritative health websites, rather than ten separate links to sift through. This improves convenience but also means reduced click-through to websites overall, as the AI has “pre-read” the content for the user. Citation Footnotes and Transparency: Unlike a raw ChatGPT response, Bing’s answers clearly highlight their sources. Small superscript numbers link to references, and users can expand a references pane or hover to see the URLs. For instance, Bing might answer a question and include “ ^{[1] [2] [3]} ” footnotes – clicking those reveals the websites (like “MayoClinic.org” or “Healthline.com”) it pulled information from ( ^[2] ). This system was lauded for bringing some credibility and accountability to AI answers ( ^[7] ). It also provides an opportunity for content publishers: if your site is one of those cited, you gain visibility and potential traffic. Dynamic Query Refinement: Users can ask follow-ups like “What about for weight loss specifically?” and Bing’s AI will remember you were asking about green tea and adapt the answer ( ^[8] ). This conversational refinement means long-tail questions that might not have a pre-written FAQ on your site could still be answered by the AI drawing from your content (if your content is comprehensive). Integrated Visuals and Features: Microsoft has enhanced Bing Chat with multimedia and interactive elements. The AI can deliver charts, graphs, or images alongside text when relevant ( ^[9] ). It can also perform certain actions like generating product comparison tables or integrating with shopping data. This blurs the line between search, content, and commerce within the chat interface. For example, an AI answer about “best smartphones under $500” might show a comparison chart of models with specs, plus links to buy – all without leaving the search page. Influence and Adoption: Bing’s move gave it a burst of attention. Within a month of launch, Microsoft announced Bing had reached 100 million daily active users , an all-time high, partly thanks to “the million+ new Bing preview users” trying the chat feature ( ^[1] ) ( ^[10] ). To put that in perspective, Google has over 1 billion daily users, but for Bing it was a significant milestone ( ^[11] ). Microsoft noted that about one-third of Bing Chat users were new to Bing , indicating the AI feature attracted people who normally defaulted to Google ( ^[12] ). On average, users were engaging in roughly 3 chats per session , with over 45 million chats conducted in the first month ( ^[12] ) – evidence that many found the conversational search useful enough to dig deeper. However, Bing’s overall search market share remains relatively small . By late 2023, estimates put Bing at around 3–4% of global search queries, barely up from pre-chat levels ( ^[13] ). For example, StatCounter showed Bing at ~3.95% worldwide vs. Google’s 89.5% in mid-2025 ( ^[14] ). In the U.S. desktop search market Bing fares a bit better (7–12% share) ( ^[15] ), but it’s still a minority player. In other words, the AI chat infusion did not instantly topple Google’s dominance in terms of market share ( ^[16] ). Many users continue Googling out of habit or because they still prefer traditional results for certain tasks. Nonetheless, for marketers, Bing Chat’s influence is bigger than its raw share numbers suggest: Precedent and Competition: Bing’s success in incorporating GPT-4 pressured Google to accelerate its own AI projects (like Bard and the Search Generative Experience, as discussed in Chapter 6). It proved the concept that AI answers with citations could be done on a large scale. Microsoft essentially made Google “dance” (as CEO Satya Nadella quipped ( ^[17] )) – forcing the 800-pound gorilla of search to respond. This opened the door for more AI search alternatives as well, creating a more fragmented search ecosystem where early adopters explore multiple tools . Traffic and SEO Impact: Bing Chat’s habit of answering questions outright means some searches that used to result in a click to a website now result in zero-click answers. Publishers have expressed concern that being merely a citation at the bottom of an AI-generated paragraph yields less traffic than being a top blue link. Indeed, users can get what they need from the summary and not click through at all ( ^[6] ). However, when users do click a Bing Chat citation, they come with high intent – they are actively seeking to validate or get more detail than the summary provided. Some websites have seen referrals from Bing’s footnotes (e.g., in their analytics, traffic from bing.com with query parameters indicating the new Bing) and consider that a new source of visits. It’s not yet massive, but if Bing grows or if those users are valuable (say, B2B researchers), it can be meaningful. New SEO Criteria – “Optimizing for the AI”: Unlike classic SEO where you optimize for an algorithmic ranking, here you’re trying to get selected and quoted by an AI . Bing has outlined some factors it uses to decide which sites to cite. These include trustworthiness and authority of the domain, content relevance to the question, clear organization and formatting, up-to-date information, and consistency of information across sources ( ^[18] ). Only a few sources will be shown for any answer (often 2–4 footnotes), making it an “exclusive club” compared to a typical SERP with 10 results ( ^[19] ). For marketers, this raises the stakes: being cited by the AI could be as valuable as ranking #1 , whereas being ignored by the AI might mean your content is practically invisible for that query. We’ll discuss optimization tactics in section 7.4, but Bing specifically has hinted that well-structured, authoritative content stands a better chance ( ^[20] ). For example, pages with clear subheadings, bullet points, FAQ schema, and concise factual statements might be easier for the AI to extract and quote ( ^[20] ). Sites that already followed best practices for Featured Snippets or Answer Boxes in Google (concise answers to common questions) have a head-start, since Bing’s AI often scoops up those same answer-friendly snippets. In summary, Bing’s GPT-4 integration has transformed its search experience into something of a hybrid between a search engine and an AI assistant. Users get quick, conversational answers with sources, and can refine through dialogue. It hasn’t dethroned Google, but it has attracted a niche of users and signaled where search is heading. Marketers should familiarize themselves with Bing Chat not only for the direct opportunity (reaching Bing’s smaller but possibly growing user base), but also because it’s a bellwether for AI-driven search norms – what works for Bing Chat (e.g. being cited for providing a clear factual answer) is likely to be similar for Google’s AI summaries and other platforms. If your content strategy adapts to win a spot in Bing’s AI answers now, you’ll also be positioning yourself well for AI features across the search spectrum . Real-World Example: Imagine a travel website that has the most authoritative content on “safest destinations for solo travelers in 2025.” In the old paradigm, you’d want to rank in the top results on Google or Bing for the query “safest places for solo travel.” In Bing Chat’s paradigm, you’d want your site to be one of the sources the AI pulls into a concise answer. If your article has a neat bullet list of “Top 10 Safest Destinations” with a brief explanation for each (and you’re a well-regarded site), Bing’s AI might respond to a user’s question with a paragraph summarizing a few top destinations and cite your list as the source ( ^[20] ) ( ^[21] ). The user gets an immediate answer (“According to TravelSafe.com and two other sources, the safest solo travel spots include Japan, New Zealand, and Iceland due to low crime rates and traveler infrastructure…”) with your site in the footnote. The user may or may not click through – but they have seen your brand as a trusted source. This kind of brand visibility via AI citation is a new outcome to optimize for.

7.2 Perplexity AI: The Rise of a Citation-Focused Answer Engine

Among the new class of AI-powered search tools, Perplexity AI has quickly emerged as an influential player. Launched in late 2022 by a San Francisco startup, Perplexity is often described as an “answer engine” or “AI search engine” that provides direct answers with source citations for virtually any question ( ^[22] ). In practice, using Perplexity feels a bit like using a supercharged Google combined with an AI assistant: you ask a question in natural language, and Perplexity returns a succinct, well-structured answer (often a few paragraphs or a list) compiled from information found on the web, with footnote numbers linking to the original sources . For example, if you ask “What are the latest trends in e-commerce for 2025?”, Perplexity might output a short summary of key trends (say, “AI-powered customer service, AR/VR shopping experiences, and sustainable packaging”) and each statement would have a tiny number like ^[1] or ^[2] that corresponds to references – perhaps a Forbes article or a market research report that it used ( ^[22] ). Clicking the footnote drops you directly into that source’s webpage (at the section relevant to the info). Focus on Trustworthy Sources: Perplexity’s philosophy is “accurate, trusted, and real-time answers” ( ^[23] ). It tries to ground every part of its answer in content from reputable websites. In essence, it is performing dozens of traditional search queries behind the scenes, analyzing the results with an LLM, and then synthesizing a coherent answer. This rigorous approach aims to reduce hallucinations and ensure that the AI isn’t just making things up – if Perplexity says something, you can inspect exactly where it got it from on the internet . Because of this design, authoritative and well-written content is more likely to be featured . Perplexity tends to cite news sites, reference sources like Wikipedia/Britannica, scholarly articles, well-known blogs, and other high-quality web pages . Low-quality or spammy sites are generally filtered out or simply not chosen by the AI as part of the answer. For marketers, this means that E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) principles remain crucial – if your site is a respected authority in its field, Perplexity is more likely to pick up your content when relevant. Conversely, thin content or SEO-gimmicky pages are unlikely to be favored by this kind of AI meta-search. Growth and Adoption: Though not (yet) a household name, Perplexity AI has seen impressive growth since its inception. By the end of 2024, the startup was valued at around $9 billion after a major funding round ( ^[24] ) – a testament to investor confidence in AI-driven search. It reportedly reached about 10–15 million active monthly users in 2024 , up sharply within a few months ( ^[24] ) ( ^[25] ). Usage statistics show surging engagement : Perplexity answered 500 million queries in all of 2023 , but then handled **250 million queries in just one month (July 2024) ( ^[26] ). That indicates a rapidly growing user base and frequency of use. In fact, by 2025 it was processing on the order of 30 million queries per day (around 780 million in May 2025) according to company statements ( ^[27] ). These numbers, while smaller than Google’s billions of daily queries, are significant for a relatively new service. For comparison, Perplexity’s website received about 73 million visits in May 2024 , which was roughly 12% of the traffic ChatGPT’s site got that month ( ^[28] ). This is notable because ChatGPT had a much bigger media profile, yet Perplexity quietly amassed a sizeable following (especially among researchers, students, and professionals seeking direct answers with citations). Demographically, Perplexity’s usage has had some interesting patterns. A large portion of its early user base came from outside the typical Western markets – for instance, a significant share of traffic has been reported from Indonesia and India, besides the US ( ^[29] ). This suggests Perplexity’s appeal as a free, reliable Q&A tool found global resonance, possibly in markets where access to high-quality information is valued and English is commonly used online. It also has mobile apps on iOS and Android, extending its reach in regions where smartphones are primary internet devices ( ^[30] ). Product Features and Differentiators: What sets Perplexity apart from both Google and even Bing Chat? A few key aspects: Always Cited, Inline:** Every factual assertion in a Perplexity answer is accompanied by a citation link. The interface often shows the answer with superscripts, and at the bottom a list of sources used. This is more akin to an academic paper or Wikipedia article style of answering. It means even a lay user can verify information immediately. Many have praised this citation-first approach as a gold standard for AI transparency ( ^[31] ). In contrast, Google’s AI summaries (SGE) initially did not cite specific sentences unless you expanded them, and ChatGPT free version doesn’t cite at all. Concise, Stackable Information: Perplexity tries to keep answers concise and to-the-point. It often aggregates multiple sources – e.g., if answering a question about climate change effects, it might pull one line from NASA, another from a UN report, etc., blending them into a single coherent paragraph. This ability to synthesize across sources (thanks to the LLM’s language skills) allows a more comprehensive answer than any single source alone. It also means the tone is neutral and informational, as it’s essentially remixing what others have said (versus injecting a distinctive AI “voice”). Users can ask follow-up questions to delve deeper or clarify, making it conversational in that sense ( ^[32] ). Real-Time and Unrestricted Web Access: Unlike some AI bots that have knowledge cutoffs or limit browsing (e.g., default ChatGPT knows little beyond 2021 data unless using plugins), Perplexity is built from the ground up to search the web. It performs live web searches when you query it ( ^[33] ), ensuring it can retrieve the latest information. If you ask “Who won the UEFA Champions League this year?” or “What is the current price of gold?”, Perplexity will go find a current source and give you the answer with today’s info, complete with the source link (which might be a news article from hours ago). This real-time capability keeps answers from going stale and makes the tool useful for newsy or time-sensitive queries (a space where static LLMs fall short). “Copilot” and Deep Research Modes: Perplexity has innovated with different modes for users. Beyond the basic Q&A interface, they introduced features like Copilot , which allows for a more interactive session where the AI will proactively ask you clarifying questions or present multiple facets of an answer. They also have a “Deep Research” mode (for paid subscribers) that can autonomously dig much deeper into a topic – performing dozens of searches and reading hundreds of sources before compiling a longer, report-like answer ( ^[33] ). This is like having a virtual research assistant gather and summarize extensive information. For example, someone doing market research can pose a broad question and get a multi-section answer complete with references. Multi-Model and Customization: Perplexity’s Pro subscription (and by 2025, perhaps even free version in some cases) allows users to pick which underlying AI model to use for generation ( ^[34] ). They’ve integrated various large language models on the backend – OpenAI’s latest GPT-4.1, Anthropic’s Claude, and even other emerging models like Meta’s Llama-based models or xAI’s Grok (the Perplexity Wikipedia entry lists those engines, implying either current or planned support) ( ^[35] ) ( ^[34] ). This is quite unique: it means the service is somewhat model-agnostic and willing to plug in the best available AI brains while maintaining the same user-facing search experience. For marketers, this doesn’t change how you optimize content, but it’s interesting in that Perplexity isn’t tied to a single AI provider – it can leverage improvements from multiple fronts (e.g., if one model is better at coding answers, another at general knowledge). Publisher Partnerships: Notably, Perplexity in mid-2024 announced a Publisher Program to share ad revenue with content creators whose pages are frequently cited ( ^[36] ). This is a very important development for the marketing and publishing community. One fear about AI answers is that they “steal” content from websites and reduce the incentive for those sites to create content (since users might not click through). By offering to share some revenue, Perplexity is acknowledging that it owes a debt to the content creators it summarizes. The details of the program aside, it signals an ecosystem approach: they want quality publishers to be on board rather than hostile. Marketers should keep an eye on such programs – it could mean that having your content cited not only yields brand exposure but potentially direct compensation in the future (a new kind of SEO monetization!). Why Perplexity Matters for GEO: For Generative Engine Optimization, Perplexity is a prime example of a platform where “content quality meets AI” . Traditional SEO practices (like keyword optimization or backlink building) matter less here than content substance and clarity . If you produce the best answer to a question, Perplexity’s AI is likely to find it and use it (regardless of whether your site is #1 on Google for that query or not). Thus, a smaller site with an excellent, well-sourced piece of content can get featured in Perplexity answers if it’s truly authoritative. From the marketer’s perspective, you’d want to ensure your content is AI-friendly : Write clear, factual statements that are easily quote-worthy (the AI might literally quote a full sentence or two from your page ( ^[21] )). Provide concise summaries or bulleted takeaways in your articles – these often get extracted. Maintain accuracy and updated info (the AI may prioritize more recent sources for current questions ( ^[37] ), so an article updated in 2025 might outrank a stale 2020 page in the AI’s eyes for a “2025 trends” query). Establish your site’s authority (through content depth and perhaps off-page signals) so it’s among the trusted pool the AI considers. Perplexity’s user base might still be modest compared to Google Search, but it includes a lot of inquisitive, high-intent users . These could be researchers, students, professionals – people looking for detailed answers. If your content gets cited, you not only gain a potential click, you gain credibility by association (“as cited on Perplexity AI” is almost like being cited in a research paper!). Some companies have observed that even when direct traffic from these AI engines is low, having their brand appear in AI-generated answers can boost brand recognition and trust . In one case study, a blog that optimized for AI answers saw its content share in AI responses rise from 2% to 12% (meaning its material was showing up a lot more in answer boxes) while traditional traffic stayed flat – indicating the benefit was mostly in audience awareness rather than clicks ( ^[38] ) ( ^[38] ). Another organization’s whitepaper became the primary cited source in many AI summaries on chatbots, which coincided with a 30% increase in demo requests and more LinkedIn mentions of their brand ( ^[39] ) ( ^[40] ). This suggests that when your content is the reference for an AI’s answer, readers implicitly value it, and that can translate to downstream actions (like seeking a demo of your product, or simply remembering your brand favorably). In conclusion, Perplexity AI represents where search is heading: ask a question, get a synthesized answer with trusted sources, and move on with confidence in the information. Its rise in 2024/2025 underscores that users are seeking efficient yet credible answers. Marketers should view Perplexity as both a testing ground for GEO techniques (since it has one of the more stringent citation mechanisms) and as an opportunity to reach a growing set of users who prefer AI-assisted search. By ensuring your content can easily be discovered and cited by Perplexity, you not only cater to that platform’s users but also prep your content to be consumable by any similar AI systems that come along.

7.3 Other AI-Driven Search Tools: DuckDuckGo, Neeva, You.com, Brave, and More

Beyond the high-profile players (Google’s AI efforts, Bing Chat, and independent upstarts like Perplexity), there’s a broader ecosystem of search engines experimenting with AI. Many of these alternatives have smaller market shares, but they often pioneer features that later become standard. They also serve specific user niches (privacy-conscious users, early tech adopters, etc.). Let’s survey a few notable ones:

7.3.1 DuckDuckGo and its AI “DuckAssist”/Duck.ai

DuckDuckGo is well-known as a privacy-focused search engine that doesn’t track users. In early 2023, DuckDuckGo jumped into the AI fray by launching DuckAssist , an AI-powered instant answer feature integrated into its search results. DuckAssist was essentially an LLM (from OpenAI/Anthropic) that would summarize information from certain sources (initially, Wikipedia and Britannica ) to directly answer user queries ( ^[41] ). It was presented as an extension of DuckDuckGo’s Instant Answers (the boxes that provide quick info, like definitions or weather) but now with AI-generated phrasing. For example, if you searched DuckDuckGo for a question like “What is the capital of Australia?”, traditionally you’d just get a snippet from a source. With DuckAssist (when it triggers), you might see a highlighted box saying: “Australia’s capital is Canberra, located in the Australian Capital Territory...” which is an AI-summarized sentence drawn from Wikipedia, accompanied by a link to the source for verification. DuckAssist did not have a full chat interface and did not take follow-up questions – it was a one-shot answer aimed at factual questions, especially those answerable by encyclopedic knowledge ( ^[42] ). DuckDuckGo explicitly mentioned this was an experiment to make search results more useful while maintaining privacy (no login required, and queries still anonymous). By March 2023 , DuckAssist rolled out to all users for relevant queries, and DuckDuckGo noted it was the first in a series of AI-assisted features ( ^[43] ). Importantly, DuckAssist would only trigger when it was pretty sure it could find a correct answer (to avoid hallucinations). It was conservative and focused on non-controversial factual questions – essentially letting the AI summarize Wikipedia, which is a high-quality source for many factual queries. The CEO, Gabriel Weinberg, even said they “fully expect it to not be perfect” but that it should help with roughly straightforward questions ( ^[44] ). Moving into 2024 and 2025, DuckDuckGo expanded on this with a broader AI strategy: They took DuckAssist out of beta and started sourcing information from across the web, not just Wikipedia ( ^[45] ). This means DuckDuckGo’s AI summaries (now often just called “AI Instant Answers” or “AI Search Answers”) can draw from multiple websites, similar to how Bing or Perplexity do, while still citing them. An illustration in DuckDuckGo’s announcements showed that the AI summary box displays tiny favicons or domain names indicating which websites were used ( ^[46] ) – reinforcing that transparency. DuckDuckGo introduced Duck.ai , an actual conversational chatbot mode accessible via their platform ( ^[47] ) ( ^[48] ). Users can click to enter a chat where they can ask follow-ups. True to its ethos, no account is needed and privacy measures are in place (DuckDuckGo routes queries in a way that masks the user’s IP from the AI providers) ( ^[49] ). Interestingly, Duck.ai allows the user to toggle between multiple models , including OpenAI and Anthropic models (GPT-4 variants and Claude) as well as open-source ones like Llama and Mistral ( ^[49] ). This multi-model approach is in line with DDG’s desire for independence – not putting all eggs in one AI basket, and giving users control. User Control on Frequency: DuckDuckGo’s implementation is very user-centric. In settings, users can choose how often they want to see AI answers: e.g., “Occasionally”, “Often”, or “Always”. Even at the highest setting, DuckDuckGo was initially only showing AI summaries for about 20% of searches (since it won’t force it if not confident) ( ^[50] ). Users can also turn it off entirely. This is different from Google’s approach where SGE, if enabled, shows up for most queries automatically. DuckDuckGo recognizes that some of its audience might be skeptical of AI or prefer classic results, so it provides that flexibility. From a market share perspective , DuckDuckGo is a minor player, but not insignificant: it handles hundreds of millions of searches per month and has steadily grown in the past decade due to its privacy proposition. As of 2025 it holds roughly 0.6–0.8% of global search market share (closer to ~2% in the US) ( ^[51] ). This is smaller than Bing or Yahoo, but its users are very loyal. Many DuckDuckGo users set it as their default for ideological reasons (avoid Google tracking) and thus may not use Google at all. That means if you ignore DuckDuckGo entirely, you might be missing a subset of users – often tech-savvy, privacy-conscious individuals (including some journalists, developers, etc.). For marketers, optimizing for DuckDuckGo’s AI isn’t radically different from general GEO principles. The engine is likely pulling answers from sources like Wikipedia (so having your brand or product represented accurately on Wikipedia can indirectly help). It also likely favors universal factual info – so if you have content you want featured, it helps if that content is referenced by known authorities or structured in a way the AI can trust. Perhaps the simplest tactic is: if there are common questions in your domain, ensure that either your site or a site that cites you provides a clear answer on Wikipedia or similar sources, because DuckAssist was heavily reliant on that. One interesting angle is community Q&A content : DuckDuckGo might not use Reddit or StackExchange as sources for DuckAssist (since it started with encyclopedias), but the general trend is that these conversational search tools do incorporate forum content (as we’ll see with You.com). It’s not far-fetched that Duck.ai’s web-wide answers might sometimes draw from forums if relevant. So participating in those discussions (Reddit, Quora, StackExchange) where appropriate can seed content that later gets summarized by various AI engines. In summary, DuckDuckGo’s venture into AI (DuckAssist and Duck.ai) shows even privacy-first companies see the value in AI summarization . They’ve implemented it in a restrained, user-friendly way – aligning with their brand (no tracking, give user choice, cite sources). While their audience is smaller, they often represent a high-value demographic and early adopters. Marketers with a focus on tech or privacy-aware consumers should ensure their SEO strategy includes DuckDuckGo (e.g., making sure their content renders well there, perhaps getting listed in DDG’s instant answers if possible). And since DuckDuckGo sources content from similar pools as others, the same optimizations that help you on Google/Bing (clear answers, structured data, authority) will help on DuckDuckGo’s AI answers too.

7.3.2 NeevaAI – An Early Innovator (Now Defunct but Influential)

Another name worth mentioning is Neeva , a startup-led search engine that was one of the first to integrate LLM-based answers with citations . Neeva was founded by ex-Google executives and launched in 2021 as an ad-free, subscription-based search alternative. In January 2023 – right on the heels of ChatGPT’s public debut – Neeva announced “NeevaAI” , a feature that would answer queries with a summarized response and cite the sources used , very much like what Bing and Google would later attempt ( ^[52] ). In fact, NeevaAI beat both Microsoft and Google to the punch – it was deployed to users in early 2023 before Bing’s February GPT-4 chat launch and while Google’s SGE was still in the lab ( ^[52] ). NeevaAI’s answers looked similar to what we see now on Perplexity or Bing: a few paragraphs answering a question, with annotations linking out. It was praised for being the “ world’s first LLM-powered answer engine with reliable citations ” by Neeva’s founders ( ^[53] ). They saw it as a way to make search more user-friendly (no ads, no clutter, just answers) while still respecting content creators (via citations). However, Neeva faced a steep uphill battle in the search market. Despite the innovative product, it struggled to attract enough users to sustain its model ( ^[54] ). People are so accustomed to free search (and to Google) that getting them to switch – let alone pay a subscription – was extremely challenging. By May 2023, Neeva announced that it would shut down its consumer search engine on June 2, 2023 ( ^[55] ). The founders cited the difficulty of breaking user habits and the tough economics (especially in an environment where generative AI itself was changing the landscape) ( ^[56] ). Indeed, as a small company, competing with Microsoft (which can afford to give away GPT-4 in Bing for free) and Google (ubiquitous default engine) was nearly impossible. Neeva’s data showed maybe ~600,000 users at peak – not enough to justify the expenses. While Neeva’s life as a search engine was short, its impact was notable: It validated the concept of integrated AI answers. Early adopters who tried NeevaAI in Jan/Feb 2023 got a glimpse of the future. In many ways, Neeva’s approach presaged what others did – and it likely spurred the giants to move faster. There’s a bit of irony that Neeva introduced a feature (AI answers) to out-compete Google, but Google and Bing’s swift responses meant Neeva lost its unique edge within months. It highlighted the importance of citations in AI answers . Neeva strongly emphasized “with citations” in its announcements, framing it as a positive differentiator. Now, citations have become a standard expectation for credible AI search (Bing does it, Google SGE does some version of it, DuckDuckGo does it, etc.). Neeva arguably helped set that standard. After shutting down search, Neeva was acquired by Snowflake (a cloud data company) in mid-2023 ( ^[57] ). The team’s expertise is being applied to enterprise AI search (like searching within business data), which is outside our scope. But it means the tech lives on elsewhere. For marketers, the direct relevance of Neeva now is minimal (since it’s offline), but there’s a takeaway: innovation can come from anywhere, but distribution is king in search . Neeva’s failure underscores that even if you have great GEO strategies, you must focus on platforms that have users. It’s a reminder to keep an eye on newcomers but also to calibrate effort vs reach. If you optimized heavily for NeevaAI early on, you may have gotten some benefit for a few months, but that effort would have been short-lived. On the other hand, if you applied those same content improvements for GEO across the board (clear answers, etc.), you would reap benefits on Bing, Google, and others too. In essence, Neeva’s story is a cautionary yet insightful footnote in the AI search saga – showing both how quickly the landscape can shift and reinforcing that Google’s dominance is hard to crack (even Bing, with all its AI splash, gained <1% market share from it ( ^[16] ); Neeva gained far less). Nevertheless, the concepts pioneered by NeevaAI (generative answers with citations) are here to stay, carried forward by those who survived. So one could say Neeva was a trailblazer that set the stage for GEO, even if it didn’t get to enjoy the spoils.

7.3.3 You.com and YouChat – The AI-Native Search Experience

Another notable player in this space is You.com , a search engine startup founded by Richard Socher (former Salesforce chief scientist) in 2020. You.com launched with the idea of a highly customizable search (users could personalize sources and apps on the results page). But its biggest pivot was going all-in on AI assistance. In December 2022 , You.com introduced YouChat , which was effectively the first search-integrated chatbot akin to ChatGPT that could handle live web queries ( ^[58] ) ( ^[59] ). This made You.com arguably the first mover in combining a conversational LLM with web search results – even before Bing did. YouChat’s capabilities: It launched using a version of GPT-3.5, tailored for conversation and equipped with real-time internet access ( ^[58] ) ( ^[59] ). This meant YouChat could answer questions about current events or recent information by fetching data from the web on the fly, then formulating an answer (with citations for the sources it used). It was essentially a more “alive” version of ChatGPT, directly embedded in a search engine interface. Early users noted that YouChat did provide source citations in its responses, although its accuracy and quality varied. The fact it could do things like summarize yesterday’s news or provide code answers with references was impressive for the time ( ^[60] ). Rapid Iterations: You.com iterated quickly on YouChat: In February 2023 , they released YouChat 2.0 , which improved conversational abilities and integrated what they called “Apps” into the chat experience ( ^[61] ). Essentially, YouChat could use specialized modules or search verticals when needed (for example, if you asked for a coding answer, it might pull from StackOverflow; if you asked about stock prices, it could pull a finance chart). They blended a custom LLM (codenamed C-A-L: Chat, Apps, Links) that could decide when to show you a chart, image, or a specific snippet from an app ( ^[61] ). This was a step toward multimodal answers – not just text, but visual elements and formatted data. It allowed users to get rich results in one interface (like a mini data table or an image result) alongside the AI’s narrative. In May 2023 , YouChat 3.0 came out, boasting deeper integration of community content and real-time info ( ^[62] ). Notably, YouChat 3.0 could directly pull content from Reddit, Stack Overflow, TikTok, Wikipedia, and more into its answers ( ^[62] ). For example, if you asked for opinions on “best programming laptop”, the AI might actually fetch a relevant Reddit thread or StackOverflow posts and incorporate insights from them (citing them). It was like having a meta-aggregator that knows where the discussions are happening. This is a clear demonstration of a trend: community-driven content often holds answers that AI can summarize . Many technical or niche questions are extensively discussed in forums; YouChat tapped into that. You.com’s approach, being a startup, was bold and experimental. It even allowed some level of user customization – you could upvote or downvote sources, and the search had tiles for different services (like a StackOverflow tile or a Twitter tile). It wasn’t just about Q&A; they also had a suite of AI “apps” (for writing help, coding, etc.), positioning themselves as an “AI hub”. User Base and Impact: You.com is much smaller in usage compared to others, likely in the low millions of queries per day range. It’s not a top search engine by market share (not in StatCounter’s top 5 globally, for instance). However, it garnered attention among the AI community and early adopters. It also likely has a significant number of users via partnerships (for instance, they launched a YouChat integration on WhatsApp to let people use AI search in messaging ( ^[63] )). For marketers, You.com might not drive noticeable traffic (unless you have a highly tech audience that you suspect uses it). But the features pioneered by You.com tell us where search experiences are heading: Real-time, on-the-fly citation of community content: If an AI can leverage Reddit or Q&A forums to answer questions, then even if your site isn’t the one being cited, the content about your product on those forums could surface. For example, if people on Reddit have discussed your product’s pros and cons and someone asks an AI “Which is better, Product X or Product Y for __?”, the AI might incorporate those opinions from Reddit (and cite the Reddit thread). Thus, your brand’s presence and reputation in online communities can directly influence AI answers . It’s a new form of off-page SEO: ensuring that experts and users speak positively and accurately about you on public forums could pay dividends when AI summarizes “what people are saying.” Structured data and app integrations: You.com’s blending of structured results (like showing a chart or a code snippet) means that providing data in structured formats (APIs, code blocks, graphs) can be advantageous. If you have an open data source or an API, an AI engine might use it to present info. For example, an e-commerce site that provides a public API for product specs might see AI search tools use that to answer “compare these two products.” The multi-turn conversation is front and center. YouChat’s existence inside a search engine affirms that search is becoming conversational . Strategies like having content in FAQ format, or anticipating follow-up questions users might ask, can align well with how these AI handle multi-turn sessions. In summary, You.com and YouChat serve as an innovation lab for AI search. While their direct reach is limited, they demonstrate how AI can integrate diverse content sources (including UGC – user generated content) and how user interaction might work (e.g., switching contexts, apps). Marketers should note the importance of community content and the fact that content from Reddit/StackExchange often ranks high in credibility for many questions. Indeed, Microsoft’s Bing AI also often pulls answers from StackOverflow for coding queries – and as a result, users can get those answers without visiting StackOverflow, a phenomenon observed and discussed in developer circles ( ^[64] ). This means companies that rely on those forums for traffic (like StackOverflow) are being disrupted, but also that if your company has domain experts , encouraging them to contribute on StackExchange or similar can increase the chances that their answers (with maybe a subtle mention of your product if allowed) get propagated by AI . Just be sure any participation is genuinely helpful and not spammy – the AI will pick up quality, not marketing fluff.

7.3.4 Brave Search and Others

A quick mention goes to Brave Search , a privacy-centric engine launched by the makers of the Brave Browser. In March 2023 , Brave Search rolled out an AI Summarizer feature ( ^[65] ). This was not a chatbot per se, but an AI-generated summary that appeared at the top of search results for certain queries. It used in-house LLM technology (not OpenAI’s) to compile a concise answer from multiple web sources ( ^[66] ). Brave explicitly designed it to always cite sources via clickable links within the summary text ( ^[67] ). For example, if you searched “What happened in the East Palestine, Ohio incident?”, Brave’s Summarizer might produce a few sentences giving the gist and include references like “[source: NYTimes][source: EPA]” as hyperlinks embedded in the summary. They bragged that unlike some AI chat tools that might fabricate, their system only pulled from actual web results and gave attribution ( ^[68] ). They also highlighted that it was done in a privacy-preserving way on their own infrastructure. Brave’s search share is small (comparable to DuckDuckGo’s range), but it’s noteworthy because they took a stance of not relying on Big Tech models – they trained their own and scaled it to handle all Brave queries (which at one point was 22 million queries per day by their claim) ( ^[68] ). For marketers, Brave’s Summarizer again reinforces the pattern: if your content has clear, authoritative info, it might get distilled into an AI summary . The good part is the user sees your link right there if they want more. The bad part is, again, they might not click if the summary suffices. But better to be cited than omitted. Other notable AI-centric search tools include: WolframAlpha , an old player (launched 2009) which is a “computational answer engine.” It’s not an LLM, but worth noting as it directly answers factual and mathematical queries from its curated knowledge base. It became relevant to AI chat when it was integrated as a plugin to ChatGPT for factual math/science answers. If you have content in the scientific or quantitative realm, WolframAlpha integration or optimization (structured data that it can ingest) might be something to consider. Kagi – a small subscription search engine that introduced its own AI summary feature called “Oracle”. It’s niche but shows even boutique engines are adding AI summaries. Ask.com (Ask Jeeves) – historically known for Q&A style, but it hasn’t been a major AI player recently. Baidu’s ERNIE Bot and Yandex’s attempt – outside English markets, players like Baidu (China) and Yandex (Russia) have launched their own LLM-based search assistants. For example, Baidu’s ERNIE Bot (launched 2023) can answer questions in Chinese in a ChatGPT-like fashion, integrated into Baidu search. While our focus is English markets, marketers targeting those regions should consider similar GEO strategies (ensuring content is accessible to those engines, possibly optimizing for how they pick sources, etc.). The overarching theme with all these alternatives is that AI-driven search is not a one-company phenomenon – it’s industry-wide . Even if some engines die off (like Neeva) or remain small, they contribute ideas that larger engines adopt. For instance, the idea of revenue-sharing with publishers started with smaller ones (Neeva talked about it, Perplexity is doing it) and now even Google is reportedly exploring ways to compensate publishers for AI usage. So, paying attention to the whole ecosystem can give you an early look at trends that might go mainstream.

7.4 Optimizing for Alternative AI Search Platforms (GEO Best Practices Across Engines)

Now that we’ve reviewed the key players beyond Google, the question for marketers is: How do we optimize for these AI-driven search experiences? The good news is that there is a lot of overlap – generally, what’s good for one tends to be good for others, because ultimately these systems all aim to provide users with accurate, authoritative answers. That said, there are some nuances. Here we’ll outline strategies to improve your visibility on Bing Chat, Perplexity, DuckDuckGo’s AI, and similar platforms.

7.4.1 Ensure Crawlability and Open Access

First and foremost, your content must be accessible to these engines. Unlike traditional search (where Google’s crawler was king), now multiple AI systems might try to read your site – Bing’s crawler, perhaps OpenAI’s browser (for ChatGPT plugins or Bing’s backend), Perplexity’s crawler, etc. Make sure you are not inadvertently blocking these. Check your robots.txt – while most AI search tools obey standard robots rules, you might consider explicitly allowing known AI user agents if applicable. Conversely, if there’s content you don’t want used by AI, that’s a separate discussion (some publishers block OpenAI’s GPTBot now). But for GEO, assume you want maximum exposure . No Paywall (for Key Content): AI answers cannot retrieve content behind logins or strict paywalls. If you run a gated content site, consider offering a summary or portion that’s publicly accessible, so AI engines have something to work with (and then can cite you, perhaps leading users to sign up for the full piece). If everything is locked away, you’ll be invisible to AI search. Some sites are experimenting with “noai” metatags to opt out of LLM training – but note that opting out might also exclude you from being cited in answers. This is a strategic choice: do you prefer not to be consumed by AI, or do you want the exposure? Most marketers will lean towards exposure, given proper credit.

7.4.2 Emphasize Authoritative, Well-Structured Content

Authority & Trust: All these AI systems heavily prioritize trusted sources . Bing’s selection criteria explicitly include Domain Authority and Content Trustworthiness ( ^[18] ). Perplexity similarly tends to quote highly credible domains. DuckDuckGo started with sources like Wikipedia. To be seen as authoritative: Build your domain’s expertise in a niche – cover topics comprehensively and accurately (depth of content). Get quality backlinks and mentions from other reputable sites (traditional SEO off-page still matters for signaling authority). Ensure your content is factual, up-to-date, and error-free. AI models cross-check facts across sources ( ^[69] ); if your data is an outlier or outdated, it might be skipped in favor of a source that has more consensus or recent info. Clear Structure & Formatting: AI answers often extract specific sentences or bullet points from content. Thus, how you format content can influence whether the AI finds it easy to use: Use Headings and Subheadings (H2, H3, etc.) that clearly delineate sections and indicate the topic of each section. For example, an FAQ page with questions as headings and answers below is excellent fodder for AI. Bullet Points and Numbered Lists are AI-friendly. If you have “Top 5 reasons” or steps in a process, list them cleanly. Bing’s AI, for instance, has been seen quoting bullet lists directly (with a citation) because it’s a concise piece of info. Schema Markup: Implement structured data like FAQ schema, How-To schema, etc. ( ^[70] ). While we don’t have direct proof that Perplexity or Bing’s AI actively parse schema, it certainly can’t hurt – and Bing’s own advice to marketers includes using schema to help it understand content ( ^[70] ). If nothing else, the presence of schema might assist the underlying search index (Bing/Google) in knowing your page has Q&A or how-to content, which could then be fed to the AI layer. Tables and Charts: Some answers might present data from tables (Brave’s Summarizer can highlight parts of a table on a page, for instance). Use tables for data comparisons where appropriate, with clear labels. Provide alt text for charts with key takeaways (since the AI might read that). Consistent Terminology: AI models pay attention to wording. If you have an important fact, phrase it in a straightforward, unambiguous way. For example, instead of burying the answer to a question in a verbose paragraph, use a declarative sentence that could stand alone . E.g., “Yes, Product X is water-resistant up to 1 meter for 30 minutes (IP67 rating).” That sentence is pluckable by an AI and could appear in an answer about “Is Product X water resistant?” – with a citation to you. Include Sourceable Quotes: This is an interesting tactic suggested for Bing ( ^[71] ). It means writing short, impactful sentences that an AI might find convenient to quote directly. Think of them as pull quotes or sound bites. E.g., “According to [Our Company] research, nearly 60% of shoppers now use AI assistants for product discovery .” If that is in your content and it’s a compelling stat (and you are a known source), an AI summary about e-commerce trends might pull that sentence and cite you ( ^[71] ). Journalists do this too when writing articles – they look for quotable lines. Now AI is doing it algorithmically.

7.4.3 Leverage Community and Off-Page Content

As noted, community forums and Q&A sites play a big role in AI answers, especially for niche and long-tail queries. Marketers should not ignore this indirect avenue: Identify key forums for your industry. It could be Reddit (there’s a subreddit for almost every topic), Stack Exchange network, Quora, or specialized forums. For developer tools, StackOverflow is king; for consumer products, maybe Reddit or Quora. Contribute value in those communities. This isn’t traditional “link building” – you’re not there to drop your link (in fact, overly promotional posts will be downvoted or removed). Instead, have genuine participation: answer questions, provide expert insights, correct misconceptions about your product or industry. Over time, some of those posts could become reference points that AI models pick up on. For example, if multiple Reddit users (including possibly your experts incognito) mention that your software has a certain feature that competitor lacks, an AI summarizing “Software X vs Y” might reflect that point, citing the Reddit thread. Encourage user reviews and Q&A on third-party sites. Google’s own generative summaries have shown content from product review sites or Q&A on e-commerce pages. Bing and others might incorporate things like Amazon’s “Most asked questions” or StackExchange answers. While you can’t directly control user-generated content, you can foster a positive presence. For instance, ensure your product’s FAQ on Amazon is answered (even if you as a brand representative answer it). The more high-quality content exists about your brand on the open web, the more likely AI finds something relevant to quote. Monitor AI outputs for your brand. Periodically, try asking these AI engines questions about your company or product: “What does [My Company] do?” or “Is [My Product] good for [use case]?” See what comes up. Does the AI mention incorrect info?

If so, trace the citation – maybe a forum post has wrong data. That gives you a clue where to jump in and clarify in the source community. This is a new kind of brand monitoring. It’s akin to checking search results for your brand, but now you check AI answers for your brand. One caveat: as of now, Bing Chat and others often will not mention a lesser-known brand in an answer unless specifically asked (to avoid sounding like endorsement). But if asked directly about you, the answer’s accuracy matters. Also, for general questions like “best project management software”, an AI might list a few tools – will yours be among them? That depends on whether the sources it draws from talk about your tool in that context. So you’d want to be present in “Top 10” list articles or discussions, etc.

7.4.4 Technical SEO and Speed – Still Important

The AI layer doesn’t eliminate the need for strong technical SEO: Site Speed and Performance: If an AI search tool is fetching your page in real-time to get info (which is how Bing operates – it fetches live content to feed GPT-4), a slow site could be a hindrance. If your page takes too long to load, maybe the engine skips it or times out. Bing’s index likely caches content, but freshness matters, so it might be hitting your site. Ensure fast response times. Also, mobile-friendliness and good UX indirectly matter since Bing and others won’t consider a poor quality page authoritative. Meta Tags and Snippets: While an AI doesn’t just parrot your meta description, having a concise meta description or snippet can influence what the search index thinks your page is about, and that context can feed into AI answer selection. Plus, Google’s SGE sometimes highlights key sentences from a page – which often come from a well-written introductory paragraph. So continue writing clear intros and use meta tags appropriately. Content Freshness Signals: Use dates on your content where applicable (and update them when you refresh an article). If Bing sees two articles on “AI search trends” and one was updated in 2025 and the other in 2022, it will likely favor the newer for a question about current trends ( ^[37] ). Perplexity also tends to fetch the latest info. This doesn’t mean old content is useless (evergreen content still works for timeless questions), but for anything time-sensitive, keep it fresh. Crawl Depth: Ensure important content isn’t buried. Traditional SEO audits (fix broken links, have good sitemaps, etc.) remain valuable because if your content isn’t indexed properly, the AI can’t use it.

7.4.5 Ethical SEO and “GEO” Tactics

It might be tempting to try to manipulate AI answers. For instance, one could think “maybe I can get my site cited by feeding misleading information somewhere.” Be very careful here. Black-hat SEO tactics in the AI era (like creating spammy Q&A pages just to get quoted, or gaming prompt answers) can backfire severely. The AI models are getting better at cross-verifying and ignoring outliers ( ^[69] ). Additionally, search engines have teams looking at the quality of AI answers – if they find people trying to trick the system, they’ll patch it (just as Google patched many SEO loopholes over the years). Instead, focus on ethical influence : Provide genuinely useful content even if the immediate click doesn’t come . This builds trust such that when an AI or human encounters your brand, they see it as helpful. For example, a financial company might publish a free dataset or tool. Perplexity’s answer might use that data – even if users don’t click, they see the company name and perhaps associate it with expertise. Avoid AI-generated content for SEO (ironically) unless you thoroughly fact-check it. Low-quality content won’t get cited by these AI engines; they’re picking the cream of the crop. So churning out 100 AI-written articles likely won’t help (and could harm your human SEO too). Quality over quantity is more crucial than ever. Mark your content clearly where needed. If you have sections that summarize others’ data, cite them (the AI might give you credit for being transparent). If you have original research, highlight that – use wording like “In our 2025 survey of 500 CEOs…” so the AI knows this is proprietary info from you. Monitor and adapt: This is new territory for everyone. Keep an eye on how these AI answers evolve. They might start showing more source links or fewer. They might change which types of content they favor as the models improve. Staying updated (via SEO news, experiment yourself, etc.) is part of GEO. In essence, optimizing for alternative AI search platforms is about making your content as AI-ready as possible : accessible, authoritative, structured, and present wherever the AI might look (your site and the broader web). The nice side-effect is that these optimizations typically improve user experience for human visitors too – clarity, structure, trustworthiness are universally good practices.

7.5 Should Marketers Care About These Alternatives? – Reach vs. Opportunity

After examining Bing, Perplexity, and others, you might wonder: given their relatively small market shares compared to Google, how much effort should you really allocate here? It’s a valid question. Google still accounts for ~90% of global search traffic ( ^[14] ); in many organizations, resources are limited, and Google SEO (plus maybe a bit of Bing SEO) has been the primary focus for years. The current reach of alternatives: Bing : ~3-4% worldwide share (a bit higher on desktop, and in certain countries like U.S./UK). Despite Bing Chat’s hype, it only nudged this a little upward (e.g., from ~3% to ~3.4% globally in 2023) ( ^[13] ). If you’re strapped for time, you might think “Why bother? That’s tiny.” However, consider that 100 million people use Bing daily ( ^[1] ). It’s not negligible. And some demographics (like Windows 11 users with the integrated Bing Chat, or corporate users on default Edge) might be heavily represented. Perplexity : not a traditional “search engine” in share stats, but 73 million visits in a month and growing ( ^[72] ). It likely has a few million regular users who ask many questions. These users might skew toward researchers, students, and professionals. DuckDuckGo : ~1% share globally ( ^[51] ), but with over 100M queries per day reported in 2022 (and likely more now). A niche but loyal base, including influencers who publicly advocate it (even Twitter’s ex-CEO Jack Dorsey has endorsed using DDG) – these users can be evangelists. You.com, Brave, etc. : collectively well below 1%, but they often attract early adopters and tech enthusiasts. These are the people who write blogs, Reddit posts, or news articles – i.e., they punch above their weight in shaping opinions. If an influential tech blogger consistently sees your site cited in Perplexity or Bing, they might mention that in an article. ChatGPT itself : Let’s not forget, some users bypass search and just use ChatGPT or other assistants to find information (“answer engines” without a traditional search engine). ChatGPT had 600 million visits in May 2024 ( ^[73] ) (though not all for search-like queries). If you have content that is embedded as part of the training data or via plugins, it could influence answers. However, with ChatGPT’s default model being a black box (no live web, no citations), optimizing for that is more about providing good data to the training corpus (which is indirect and not something you do on the fly). That’s a complex topic, likely touched on in Chapter 12 (Prompt Optimization & Ethical Influence). Given the above, **marketers should care strategically : Early Movers Advantage:** The fact that these platforms are smaller actually means less competition in them right now for GEO. Many businesses have not yet systematically tried to optimize for being cited in AI answers. By starting now, you can carve out a presence. For example, maybe few of your competitors have bothered to ensure their FAQ content is in Q&A form and well-cited – if you do it first, Bing’s AI might start preferring your site for answers in your sector. Capturing visibility on Bing Chat or Perplexity in 2024 could be analogous to capturing top Google ranks in the early 2000s – easier before everyone else piles on. Influential Audience: As mentioned, early adopters of AI search skew tech-savvy, and often influencers or decision-makers . Think of developers (using Bing Chat integrated in VSCode or StackOverflow discussions), journalists (experimenting with Perplexity to gather facts for an article and seeing your brand cited), or executives (asking ChatGPT/Bing for quick insights). If your site is frequently cited in these AI-generated answers, it subconsciously builds credibility with these users. Even if they don’t click immediately, your brand may be seen as a go-to authority. Then, when they encounter your brand elsewhere, there’s recognition. It’s akin to brand impressions in display advertising – sometimes you’re doing GEO for the branding as much as the click. Cross-Pollination to Google: Many GEO tactics for Bing/Perplexity will help with Google’s Search Generative Experience too. Google’s SGE, while outside this chapter’s scope, also presents AI summaries with cited links (in labs as of 2023). The way to get cited by Google’s AI is very much to have the same qualities: authoritative, concise content that directly answers questions. So by optimizing for the alternatives, you are indirectly preparing for Google’s imminent AI-dominated results. It’s a form of future-proofing your SEO. As one SEO expert put it, “view GEO as an extension of SEO rather than a replacement” ( ^[74] ) – the core idea is still to produce great content; we’re just tuning it for AI consumption. Incremental Gains: Even if Bing only brings, say, 5% of your search traffic, that might still be significant in absolute terms. If your site gets 1,000,000 visits from Google a month, an extra 50,000 from Bing is not trivial – especially if competition for those users is lighter, conversion rates could be higher (maybe Bing users are less inundated with options). Similarly, a handful of mentions on Perplexity that lead researchers to cite your study, or a few thousand DuckDuckGo users who happen to be highly relevant (e.g., privacy enthusiasts who might love your privacy-friendly product) – those are valuable. Marketers often chase every bit of market share; ignoring even a 5-10% segment entirely would be a missed opportunity. Case Study – B2B Tech: Consider a B2B software company targeting developers. Traditional SEO might bring in organic traffic via Google, but many developers are now using Stack Overflow less and tools like GitHub Copilot or Bing Chat more ( ^[64] ). If your product documentation is top-notch and accessible, a dev might ask Bing Chat a question and get an answer citing your docs – saving them the trip to StackOverflow and directly building trust in your product’s resources. That developer may not click right then (they got the answer), but next time they have a deeper issue, they might recall “that answer was from Company X’s docs, maybe I should check their site.” Moreover, we saw earlier how Stack Overflow’s traffic has dipped partly because AI is serving answers directly ( ^[64] ). If competitors relied solely on community answers for visibility, and you instead ensure your official content is good enough to be used by AI, you could leapfrog in visibility. All that said, prioritization is key . The advice isn’t to abandon Google optimization in favor of Bing or others, but rather to incorporate GEO strategies into your overall search marketing plan . It might mean reallocating some content resources to update FAQ pages for AI-friendliness, or spending a bit of time on Bing Webmaster Tools (yes, Bing has its own) to see how you’re performing there. It could involve monitoring where your content is being cited by these AI engines – which itself could become a new metric (perhaps “AI Citation Count” will be a KPI in future SEO reports, analogous to search impressions). Looking ahead, it’s quite possible that alternative engines could grow . For example, if Microsoft continues to invest (it’s leveraging Bing Chat across Windows, Office, and more), Bing’s share could inch upward, or at least **Bing’s influence might extend beyond bing.com (e.g., Bing answers powering things in other apps). Similarly, Apple has been rumored to be working on AI search capabilities – if they launch something on iPhones, that could be a new channel overnight. By getting experience with optimizing for Bing/Perplexity, you’ll be better prepared for any newcomer. In conclusion, while Google remains the primary focus, forward-thinking marketers should absolutely care about the AI search alternatives . The risks of ignoring them include missing out on early adopter audiences, ceding mindshare to competitors who do engage there, and being unprepared when Google fully rolls out similar paradigms. On the flip side, the benefits of engagement include enhanced brand authority (through citations), additional traffic (even if modest, it can have high ROI if it’s cheap to get), and crucial learning that can inform your overall content strategy. As one LinkedIn analysis noted, neglecting GEO now could relegate your content to a mere footnote in others’ AI responses** ( ^[75] ) – meaning you might only be referenced as supporting someone else’s more prominently featured info. Better to strive to be the primary source being quoted. By capturing visibility on platforms like Bing Chat and Perplexity today, you not only gain an edge in those arenas, but you also refine the skills and content quality needed to thrive in the evolving world of AI-assisted search. In the next (and final) chapters, we’ll look at measuring GEO success and preparing for future trends – but as far as where to optimize, remember that Google may be king, but it’s now ruling a more crowded kingdom . The savvy marketer will build presence in all the places the audience seeks answers, however small, because those small streams can collectively turn into a significant river of opportunity – and they often carry the most influential currents. Key Takeaway: Alternative AI search engines might individually have modest reach, but they attract influential early adopters and pioneer features that larger engines emulate. Marketers should proactively optimize for these platforms – it’s largely an extension of good SEO/GEO practices – to gain brand visibility and be ahead of the curve. The effort is justified not just by the direct traffic available now, but by the strategic advantage of positioning your content for an AI-driven search future where being the trusted cited source is gold. ( ^[39] ) ( ^[64] )

References

^[1] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2023/3/9/23631912/microsoft-bing-100-million-daily-active-users-milestone

^[2] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[3] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[4] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[5] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[6] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[7] Becomingthemuse.Net Article - Becomingthemuse.Net URL: https://becomingthemuse.net/2023/03/01/ai-powered-bing-chat

^[8] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[9] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[10] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2023/3/9/23631912/microsoft-bing-100-million-daily-active-users-milestone

^[11] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2023/3/9/23631912/microsoft-bing-100-million-daily-active-users-milestone

^[12] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2023/3/9/23631912/microsoft-bing-100-million-daily-active-users-milestone

^[13] www.pymnts.com - Pymnts.Com URL: https://www.pymnts.com/artificial-intelligence-2/2024/report-chatgpt-hasnt-helped-bing-compete-with-google

^[14] Gs.Statcounter.Com Article - Gs.Statcounter.Com URL: https://gs.statcounter.com/search-engine-market-share

^[15] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/one-year-later-little-change-to-microsoft-bing-search-market-share-437238

^[16] Arstechnica.Com Article - Arstechnica.Com URL: https://arstechnica.com/ai/2024/01/report-microsofts-ai-infusion-hasnt-helped-bing-take-share-from-google

^[17] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2023/3/9/23631912/microsoft-bing-100-million-daily-active-users-milestone

^[18] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[19] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[20] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[21] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[22] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[23] www.perplexity.ai - Perplexity.Ai URL: https://www.perplexity.ai

^[24] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[25] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[26] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[27] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[28] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[29] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[30] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[31] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/googles-agent2agent-protocol-2566

^[32] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[33] www.perplexity.ai - Perplexity.Ai URL: https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research

^[34] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[35] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[36] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/Perplexity_AI

^[37] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[38] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/generative-engine-optimization-geo-vs-traditional-seo-francis-dba-b01ic

^[39] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/generative-engine-optimization-geo-vs-traditional-seo-francis-dba-b01ic

^[40] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/generative-engine-optimization-geo-vs-traditional-seo-francis-dba-b01ic

^[41] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/DuckDuckGo

^[42] Voicebot.Ai Article - Voicebot.Ai URL: https://voicebot.ai/2023/03/08/new-duckduckgo-generative-ai-feature-summarizes-wikipedia-articles-for-instant-answers

^[43] Spreadprivacy.Com Article - Spreadprivacy.Com URL: https://spreadprivacy.com/duckassist-launch

^[44] Arstechnica.Com Article - Arstechnica.Com URL: https://arstechnica.com/information-technology/2023/03/wikipedia-ai-truth-duckduckgo-hopes-so-with-new-answerbot

^[45] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[46] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[47] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[48] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[49] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[50] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/624899/duckduckgo-ai-search-chatbot-plans

^[51] Gs.Statcounter.Com Article - Gs.Statcounter.Com URL: https://gs.statcounter.com/search-engine-market-share

^[52] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[53] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[54] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[55] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[56] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[57] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/neeva-shutting-down-427384

^[58] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/You.com

^[59] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/You.com

^[60] Synthedia.Substack.Com Article - Synthedia.Substack.Com URL: https://synthedia.substack.com/p/youchat-is-like-chatgpt-with-real

^[61] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/You.com

^[62] En.Wikipedia.Org Article - En.Wikipedia.Org URL: https://en.wikipedia.org/wiki/You.com

^[63] You.Com Article - You.Com URL: https://you.com/articles/you.com-introduces-ai-powered-search-on-whatsapp

^[64] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/decline-stack-overflow-age-ai-alternatives-julian-van-dijk-uwlne

^[65] Brave.Com Article - Brave.Com URL: https://brave.com/blog/ai-summarizer

^[66] Brave.Com Article - Brave.Com URL: https://brave.com/blog/ai-summarizer

^[67] Brave.Com Article - Brave.Com URL: https://brave.com/blog/ai-summarizer

^[68] Brave.Com Article - Brave.Com URL: https://brave.com/blog/ai-summarizer

^[69] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[70] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[71] Blueinteractiveagency.Com Article - Blueinteractiveagency.Com URL: https://blueinteractiveagency.com/seo-blog/2025/06/integrate-bing-chatgpt-search-functions

^[72] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[73] Electroiq.Com Article - Electroiq.Com URL: https://electroiq.com/stats/perplexity-ai-statistics

^[74] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/generative-engine-optimization-geo-vs-traditional-seo-francis-dba-b01ic

^[75] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/generative-engine-optimization-geo-vs-traditional-seo-francis-dba-b01ic

8. Emerging LLMs and Open-Source Models – Claude, LLaMA, and Grok

Chapter Overview: In this chapter, we explore the rise of new large language models (LLMs) beyond the early leaders like OpenAI’s GPT-4. We examine Anthropic’s Claude (known for its large context window and safety measures), Meta’s LLaMA (an open-source model family powering countless niche applications), and xAI’s Grok (Elon Musk’s social-media-trained chatbot with an irreverent style). We compare their key features – from context length and knowledge updates to integration channels – and discuss what these differences mean for Generative Engine Optimization (GEO). Finally, we consider how a multi-model ecosystem is emerging, where no single AI assistant dominates all queries, requiring marketers to optimize content for a variety of AI systems globally.

8.1 Anthropic’s Claude: Long Memory and Safety-Focused Design

Anthropic’s Claude is often mentioned in the same breath as GPT-4, positioned as a major competitor in advanced AI chatbots. Claude’s defining feature is its massive “memory” or context window , which allows it to read and retain extremely large amounts of text in one conversation. In mid-2023, Anthropic expanded Claude’s context window from an already-impressive 9,000 tokens to 100,000 tokens , roughly 75,000 words ( ^[1] ). This means Claude can ingest entire books or lengthy documents at once without losing track. For example, Anthropic demonstrated Claude reading The Great Gatsby (72K tokens) and correctly identifying a single modified line in just 22 seconds ( ^[2] ). Such capability far exceeds the context limits of most other models at the time, enabling deep analysis of long-form content in one go. In fact, newer Claude versions (Claude 2.1 and “Claude 4” in 2024–25) have pushed context limits even further – reportedly up to 200K tokens (around 150,000 words) in some variants ( ^[3] ) ( ^[4] ) – making Claude especially suited for tasks like reviewing lengthy reports, synthesizing information across multiple files, or digesting entire websites’ content in one answer. Claude is not just about length; it’s also designed with a strong emphasis on safety and transparency . Anthropic developed Claude using a technique called “Constitutional AI,” which means the model follows a set of guiding principles (a kind of internal AI constitution) to ensure it produces helpful and harmless responses ( ^[5] ). This approach makes Claude more cautious and polite in tone, avoiding toxic or disallowed content more rigorously. For enterprise users and marketers, this reliability is a selling point – Claude aims to minimize the risk of offensive or wildly incorrect outputs through these built-in safety rules. Anthropic often touts this as a differentiator, appealing to businesses that require an AI assistant aligned with ethical guidelines and brand safety. Another strength of Claude is how it’s being integrated into real-world productivity tools , enhancing its practical utility. For instance, Slack’s workplace messaging platform has incorporated Claude as an AI assistant for teams. In Slack, users can @mention Claude to summarize long chat threads, answer questions, or pull information from documents and websites shared in the channel ( ^[6] ). Because of Claude’s large memory, it can remember an entire Slack conversation history, meaning it can provide context-aware answers even in extended discussions. Notably, Claude can also fetch content from a URL you share (when explicitly asked), allowing it to include up-to-date information from the web in its responses ( ^[6] ) ( ^[7] ). However, Claude’s core knowledge (from its training data) has a cutoff of roughly 2021–2022, so by itself it “has not read recent content” and “does not know today’s date or current events” ( ^[7] ). The Slack integration mitigates this by letting Claude read provided links or user-supplied text, effectively performing on-demand retrieval. Anthropic assures Slack users that any data Claude sees in your workspace is kept private and not used to further train the model ( ^[8] ), addressing data security concerns for companies. This use case – AI assistance in corporate chat – plays to Claude’s strengths in summarization and following instructions across lots of text, like policy documents or project notes, within a controlled environment. Claude has also been made available through Quora’s AI chatbot hub called Poe . Poe is a platform that offers access to multiple AI models (GPT-4, GPT-3.5, Claude, etc.) in one app, allowing users to converse with each and even compare answers. Quora’s team, in partnership with Anthropic and Google Cloud, deployed Claude on Poe to great effect – millions of users’ questions are answered by Claude daily on that platform ( ^[9] ). According to Quora’s Product Lead for Poe, Claude delights users with its “intelligence, versatility, and human-like conversational abilities,” powering a broad range of queries from coding help to creative writing ( ^[9] ). Notably, Quora leverages Claude’s strength by using it for complex tasks like generating interactive app previews and even coding assistance within Poe ( ^[10] ). The fact that Quora felt the need for a multi-model approach in Poe (offering Claude alongside OpenAI and other models) underscores how Claude provides unique value – often excelling at detailed, structured answers and large-scale analysis – that complements other chatbots ( ^[11] ). From a GEO (Generative Engine Optimization) perspective, Claude’s rise means that content creators should be mindful that Claude can consume very large chunks of content in one go . If a user query on Slack or Poe triggers Claude to analyze “your entire report” or a 50-page whitepaper on your site, Claude can actually do it – and quickly. This raises interesting opportunities: well-structured, comprehensive content might be fully digested by Claude, potentially allowing more nuanced answers to incorporate your details. For example, a marketer could upload a lengthy product manual or a series of blog posts into a Claude-powered system and get a synthesized answer that weaves together points from all of them. If your content is rich and authoritative, Claude might present a thorough summary of it (though caution: it may not cite you unless the interface is designed for that). The flip side is that if a competitor has more concise, well-organized summaries , Claude might favor those summaries internally when formulating an answer because it can easily traverse hundreds of pages. Thus, ensuring that important information is not buried too deep in a bloated document can help – even though Claude can read it all, you want your key points to stand out clearly for any AI summarization. Claude’s safety-first orientation also implies that “black-hat” manipulative tactics are likely to backfire. Its Constitutional AI might refuse to output content that seems biased or promotional beyond factual referencing, especially if it conflicts with its principles. Marketers should therefore focus on factual, helpful content because Claude will often avoid overly promotional language or unsupported claims in the answers it generates. In practical terms, if you attempt to game the system by stuffing a page with certain phrases, Claude’s interpretation (unlike a search engine’s ranking algorithm) will be to simply summarize or analyze the content’s actual substance. It won’t be swayed by keyword density or meta tags – it “reads” like a human would. This reinforces the importance of clarity and E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) in content: Claude will pick up on expertise indicators (like citing data, providing nuanced explanations) and likely respond in kind, versus parroting marketing fluff. In summary, Anthropic’s Claude has carved a niche as the high-memory, high-integrity AI assistant . It’s being used in professional contexts (Slack for internal knowledge, Poe for broad Q&A) where its ability to swallow whole libraries of text and adhere to ethical guidelines is valued. For GEO, the emergence of Claude means content that is long-form and high-quality can shine – especially if it’s the kind of deep material that a user might feed into an AI for analysis (e.g. research papers, detailed guides). Optimizing for Claude isn’t about technical SEO tweaks, but rather about making substantive content available and easy to parse . Ensure your full guides, reports, and FAQs are accessible (no paywalls or robots.txt blocking Anthropic’s partners like search indexers) so that if a question arises, an AI like Claude can actually retrieve and absorb your content when needed. As we’ll see, Claude’s open-book, long-memory approach is a contrast to some other models – and it exemplifies how diverse LLM designs are influencing optimization strategies.

8.2 Meta’s LLaMA and the Open-Source Wave

One of the most significant developments in the AI world was Meta’s decision to release LLaMA (Large Language Model Meta AI) as an open model. In July 2023, Meta introduced LLaMA 2 as a freely available LLM for both research and commercial use, effectively open-sourcing a top-tier model ( ^[12] ) ( ^[13] ). This marked a turning point: while models like GPT-4 remained proprietary, LLaMA 2’s weights were downloadable, meaning anyone could run the model on their own hardware or fine-tune it to create a customized chatbot. Meta explicitly framed this as an open innovation approach, arguing that broad access would spur experimentation and make AI better (and safer) through community oversight ( ^[14] ) ( ^[15] ). In the first iteration (LLaMA 1), they had over 100,000 requests from researchers to access it, and many “built on top of it” to create new applications ( ^[16] ). With LLaMA 2, Meta went further – partnering with cloud providers like Microsoft Azure and Amazon AWS to host the model, and optimizing it to even run on a local PC (Windows) for developers ( ^[13] ). In short, Meta effectively donated a powerful engine to the public, betting that widespread adoption would establish LLaMA as a foundation for the next generation of AI apps. This bet seems to be paying off, as we’ve witnessed an open-source LLM boom in 2024–2025. By mid-2024, Meta reported that their LLMs (LLaMA 1, 2, and even early LLaMA 3 versions) had been downloaded over 170 million times ( ^[17] ). A vibrant ecosystem of developers sprang up to fine-tune LLaMA for specific domains and languages. Unlike closed models tied to one interface (e.g. ChatGPT to OpenAI’s chat or Bard to Google’s), open LLMs like LLaMA can be embedded anywhere . Companies across industries have taken these models and tailored them to niche applications : Education example: A South Korean startup, Mathpresso, fine-tuned LLaMA 2 to create “MathGPT,” a math tutoring chatbot used in 50 countries ( ^[18] ). They chose LLaMA over an API like ChatGPT because they needed deep customization – aligning the AI with specific curricula, exam styles, and local teaching methods ( ^[19] ). The result was an AI that could handle local educational nuances and even set world records on math problem benchmarks ( ^[20] ). Mathpresso’s co-founder noted that off-the-shelf models lacked the needed customization for complex educational needs ( ^[21] ), whereas with LLaMA 2 they could integrate their own data and expertise. This illustrates how open models enable industry-specific optimizations that would be hard to achieve with a one-size-fits-all chatbot. Business software example: Zoom, the video conferencing giant, incorporated LLaMA 2 (alongside other models) to power its Zoom AI Companion features ( ^[22] ). This assistant can summarize meetings, highlight action items, and draft chat responses – essentially acting like a smart secretary for your virtual meetings. By leveraging an open model, Zoom could integrate the AI within its own application , ensure data privacy (since they can self-host the model or use a preferred cloud), and tweak the model for the formal language and context of business meetings. It shows that even large enterprises sometimes opt for open models to build in-house AI features instead of relying solely on external APIs. Medical domain example: Researchers from EPFL and Yale took LLaMA 2 and created “Meditron,” a specialized medical AI assistant ( ^[23] ). They compressed vast medical knowledge into a conversational tool that can help with diagnoses, aimed at low-resource healthcare settings. Impressively, when Meta released an updated LLaMA 3 model, the team fine-tuned the new version within 24 hours to produce a better Meditron bot ( ^[24] ). This agility – updating a domain-specific AI in a day – highlights the advantage of having direct access to model weights. For marketers in healthcare or other regulated industries, it hints at a future where custom AIs trained on your proprietary content become part of your product or customer service. If you run a medical portal, for instance, an open model could be fine-tuned on your articles and patient FAQs to create a virtual health assistant. That assistant might never directly cite your site in an answer, but it’s essentially powered by your content behind the scenes. The “open-source wave” extends beyond Meta’s models. Numerous organizations globally have released high-quality LLMs under permissive licenses. For example, Mistral 7B (from a French startup) is a 7-billion-parameter open model launched in late 2023 that outperformed some 13B+ parameter models (like LLaMA-13B) on benchmarks ( ^[25] ), showing how smaller open models can be very efficient. The UAE’s Falcon 40B (by the Technology Innovation Institute) was another top-rated open model in 2023, made freely available and rivaling the best closed models at the time in certain tasks. Even regional efforts are underway: in Japan, telecom company NTT unveiled an LLM called “tsuzumi” in 2024, designed for Japanese language excellence and lightweight enough to run on a single GPU ( ^[26] ). Tsuzumi aims to excel at Japanese tasks and indicates how countries/companies are creating their own models for data sovereignty and local needs. In China, while OpenAI and others are restricted, companies like Baidu and Alibaba have their own models (ERNIE, Tongyi Qianwen, etc.), and some Chinese open models have been released for local use. In essence, there’s now a global multitude of LLMs, many of which are open or semi-open. For content marketers, this proliferation means your content can surface in unexpected places . Open LLMs don’t have a single “search engine” front-end. Instead, they might be built into apps, IoT devices, enterprise software, or novel search tools. Your SEO strategy can’t stop at “will Google rank this?” – you should also ask “could an AI developer use my content to train their model or feed their chatbot?” In practical terms: Publicly available content becomes training fodder. If your website has a permissive crawl policy (no restrictions) and is rich in a certain domain, you may find open-model enthusiasts fine-tuning a model on it. For example, a fintech company might train a small LLM on all SEC filings and major finance blogs (including yours) to build a finance QA bot. That bot’s answers will contain insights from your content, but users may never visit your site or even know the source. This is both a risk (losing traffic/attribution) and an opportunity – your expertise still reaches the audience indirectly. To balance this, some companies choose to publish data sets or libraries explicitly so that if models are trained, at least the source is acknowledged. Others might embed subtle references or unique phrasing in content that, if echoed by an AI, could signal that it came from them (almost like an informational watermark). Structured data and open licenses can amplify your presence. Consider releasing some content under a Creative Commons license or providing a dataset version of your content. Open-source LLM projects often grab content from sites like Wikipedia, StackExchange, or Common Crawl (a web scrape corpus). Ensuring your content is accessible to these crawlers – and ideally included in respected open data sources – increases the likelihood it’s part of the model’s knowledge. Some organizations feed their FAQs into Wikipedia or contribute expert information to Wikidata, knowing that many models ingest those sources. This way, even if your site isn’t directly scraped, your information lives in the training data stew. Monitoring and feedback loops : It becomes crucial to monitor how open models represent your brand or facts. Since anyone can spin up a LLaMA-based bot, you might find dozens of variants answering questions about your industry. Some might be outdated or fine-tuned on biased data. Unlike with Google Search (where you at least see how you rank), with open AI outputs you may need to proactively test queries on popular open-source model demos or communities (many AI forums discuss model outputs). If you find inaccuracies, it may require reaching out to the model creators or publishing clarifications widely so that future versions get the correct info. One example of content surfacing in new ways is the Perplexity AI search engine (which we discuss in Chapter 7). It uses an open or hybrid model underneath and crawls the web in real-time, then presents answers with citations. If you’ve optimized for traditional SEO, you might unknowingly be doing GEO for Perplexity too – since it will pull directly from your webpage if it has the best answer and then cite you. As more startups build vertical-specific QA bots (imagine an AI legal advisor that’s trained on open legal databases and major law firm blogs), having your content present and well-structured in those databases is key. Crucially, open models decentralize the playing field . Google’s and OpenAI’s dominance is challenged by thousands of smaller AIs that collectively reach millions of users. Meta noted that LLaMA models are being used in fields like customer service and medicine already ( ^[17] ). For marketers, this means that improving general web visibility (through SEO, PR, and being part of public knowledge sources) feeds into GEO indirectly. Your content’s journey might be: published on your site → scraped into an open dataset → fine-tuned into a niche model → deployed in an app → provides an answer to an end-user’s query. That user might never see your site, but the quality and correctness of your content still matters immensely because it influences the model’s output. If your content is incorrect or thin, an AI answer built on it will also be flawed – which could reflect poorly on your brand if noticed. Conversely, if your content is uniquely insightful (e.g., original research or expert opinions), even an uncited AI answer may prompt users to wonder “where did that come from?” – potentially leading them to search for the source (some savvy users do reverse-text searches of AI answers to find the original). In summary, Meta’s LLaMA opened the floodgates for AI everywhere . The open-source LLM wave empowers organizations to roll out their own chatbots, which means your SEO content can show up anywhere an LLM is used – not just on search engine results pages. To ride this wave, ensure your content is not only optimized for search ranking but is also readable and parsable by AI , shared in open formats, and injected into the streams of data that feed these models. Embrace the idea that any text you publish might be read by an AI and integrated into its brain. This might sound daunting, but it all circles back to producing high-quality, original content and making it accessible. Do that, and you increase the chances your brand’s knowledge will permeate through the new open AI ecosystem, even if the user doesn’t come through your front door.

8.3 xAI’s Grok: A Social Media-Fueled Challenger

No discussion of emerging LLMs would be complete without Grok , the buzzy new entrant from Elon Musk’s AI startup, xAI. Launched in late 2023 amid much fanfare, Grok has positioned itself as a maverick alternative to ChatGPT and Claude, with Musk touting it as a more irreverent, truth-seeking AI companion. In Musk’s own words, Grok is meant to be a “ politically incorrect ” chatbot – essentially a rebuttal to what he calls “woke” AI from Silicon Valley ( ^[27] ). This edgy positioning is more than just marketing; it’s reflected in Grok’s design, training data, and behavior, all of which have implications for how brands might be discussed by such a model. One of Grok’s unique angles is its deep integration with real-time social media data . Specifically, Grok has access to the stream of posts on X (formerly Twitter) in a way other chatbots do not ( ^[28] ). It can pull the latest public posts and trends from X to inform its answers. In practical terms, this means Grok is exceptionally up-to-date on current events and internet chatter . Ask Grok about a breaking news story or the day’s viral meme, and it can respond with information (and even opinions) gleaned from X posts just minutes or hours old. This real-time awareness gives it an edge in immediacy over models like ChatGPT, which rely on either static training data or slower retrieval plugins. As one early user noted, “One of the main advantages of Grok is its real-time access to X posts, which allows it to provide up-to-date information on current events” ( ^[28] ). For marketers, this feature is a double-edged sword. On one hand, if your brand is trending on social media – say, you launched a new product that everyone’s tweeting about – Grok could pick that up and mention those fresh reactions or facts in answers. On the other hand, if misinformation or negative sentiments about your brand are spreading on X, Grok might also reflect those, potentially amplifying a PR issue through its responses. Grok’s training also includes other public web data, but Musk has hinted that the X data firehose is a key differentiator. It’s as if Grok has one ear constantly to the ground of public opinion. Beyond just factual updates, Grok’s creators encourage it to have a bit of “personality” and humor . The chatbot is explicitly allowed (even encouraged) to crack jokes, use casual language, and not shy away from controversial topics. It reportedly has a “rebellious streak” and a mode where it will respond with witty insults or vulgar jokes if prompted ( ^[29] ). In regular mode, it’s more toned down, but still less filtered than say, ChatGPT. This ethos of Grok manifests in it sometimes giving answers that other AIs would refuse. For example, early testers found Grok would comment on politically sensitive queries or edgy humor that other bots would usually avoid. Musk’s vision was for an AI that pushes boundaries , and indeed Grok’s launch version did push them – perhaps too far at times. By mid-2024, Grok had gone through a couple of iterations (Grok 1.0, 1.5, 2.0…) and each time it ramped up capabilities. As of early 2025, xAI had released Grok 3 , which Musk boldly labeled “the smartest AI on Earth” ( ^[30] ) (an obvious hyperbole, but it indicates confidence). Grok 3 improved its reasoning and knowledge, and in a July 2025 livestream Musk unveiled Grok 4 , showcasing advanced problem-solving (even generating an image of colliding black holes on the fly) ( ^[31] ). They demonstrated Grok 4 solving graduate-level math problems and teased its dominance on certain AI benchmarks ( ^[32] ). Clearly, xAI is racing to match or exceed the technical prowess of GPT-4 and Claude, not just be a novelty. Grok is offered as a subscription service (around $30/month for standard access, with a pricier $300/month “Heavy” plan for power users or enterprises) ( ^[33] ), and notably, Musk announced it would be integrated into Tesla vehicles as soon as late 2025 ( ^[34] ). This means drivers (or passengers) could ask their car’s AI a question and get Grok’s answer via the car interface. It’s an intriguing distribution model – leveraging Musk’s ecosystem (X platform, Teslas) to gain users. From an optimization standpoint, how might Grok handle brand or product queries? Given its style, Grok might respond with a mix of factual info and sardonic commentary. For example, ask Grok “What do people think of Brand X’s latest phone?” It might pull in actual tweet sentiments (“I’m seeing mixed reactions on X – some users love the new camera, others say it overheats”) and then perhaps add a quip like “In other words, a typical day in tech launches 😜.” If Brand X had a recent controversy on social media, Grok could surface that too (“Also, there’s a meme about the CEO’s comment that’s trending”). This kind of answer is very different from the neutral, measured tone of a Bard or Bing AI answer, which might just list specs and reviews. Marketers should be prepared for the fact that Grok will not necessarily present your carefully crafted messaging – it will present what the crowd is saying, possibly unfiltered. In one incident, Grok even produced content containing anti-Semitic tropes because it picked up on some extremist prompts or posts ( ^[35] ), leading to public backlash and xAI removing those outputs. Musk acknowledged Grok was “too eager to please” and would answer even inappropriate prompts, necessitating dialing it back a bit ( ^[36] ). This incident is a caution: Grok’s openness can veer into offensiveness or misinformation more easily than other AIs, because its guardrails were initially looser. xAI has since adjusted some safety settings (they don’t want a PR disaster undermining the whole project), but Grok still remains comparatively less constrained than its peers. So what does this mean for those looking to optimize content for or against Grok? First, monitor social media sentiment closely – it’s effectively part of the SEO (or GEO) work now. If you launch a campaign and it’s blowing up on X, that’s not just a social media concern; an AI like Grok could propagate those reactions to users who weren’t even on X. Conversely, building a strong positive presence on X can directly influence Grok’s knowledge. For instance, providing timely, helpful answers via your official Twitter (X) account could seed Grok with expert information. Because if someone asks Grok a related question, it might recall “there was a detailed thread by @YourBrand that got a lot of engagement” and summarize points from it. Ensuring your brand’s tweets are informative and not just promotional might make the difference between Grok portraying you as an authority versus ignoring you. Second, consider engaging with xAI’s platform directly if possible. As of now, xAI hasn’t announced plugin support or a formal way to feed data to Grok beyond the public web. However, given Musk’s focus on community (and perhaps given that xAI is smaller than OpenAI), they might allow user submissions or custom data integration in the future. For example, if xAI opens an API or a business partnership program, being early to experiment could pay off. Imagine being the first retail company to integrate your product catalog into Grok’s knowledge – when users ask Grok for gift suggestions, it might draw from your up-to-date catalog instead of older data. Third, brace for Grok’s style . If you find Grok mentioning your brand in a snarky or humorous way, it might be futile to “correct” the AI (since it’s behaving as designed). Instead, incorporate that into your marketing thinking. Some brands might even enjoy the edgier mentions – it can humanize a corporate image if an AI jokes about it in good spirit. For example, if Grok quips “Brand Y’s new sneakers are so popular even aliens on Mars want a pair (per Elon’s other company 🚀),” a clever social media manager could riff on that joke. In contrast, false or damaging claims need addressing at the root: which likely means dispelling the rumor on social channels or through press so that the chatter dies down and Grok’s real-time feed moves on. Finally, Grok underscores that the AI landscape is diversifying in tone and audience . While ChatGPT might skew towards professional and educational uses, Grok is clearly aimed at a more casual, perhaps younger or internet-savvy crowd (think meme enthusiasts, crypto bros, or those who frequent Musk’s circles). If that overlaps with your target audience, you have to pay attention to Grok. It’s not yet as widely used – estimates put its user base far below ChatGPT’s; StatCounter’s data of mid-2025, for instance, didn’t even list Grok by name in global chatbot market share (implying it was under 1%) ( ^[37] ) – but with integration into X and Tesla, its reach could quickly grow. Musk’s ventures have a way of cross-pollinating users (imagine every Tesla driver gets curious and tries Grok – that’s millions of potential users overnight). In summary, xAI’s Grok represents an unorthodox but important new channel. It combines the pulse of social media with the capabilities of a modern LLM, wrapped in a provocative persona. For marketers, the rise of Grok means social media management and GEO intersect more than ever. Ensuring your brand’s narrative on platforms like X is accurate and engaging isn’t just for the humans scrolling feeds – it’s for the AIs like Grok that are listening in and will retell that story to others. Embrace Grok’s existence as both a challenge (it might say things out of your control) and an opportunity (a chance for your brand to be part of real-time cultural conversations facilitated by AI). And above all, keep an eye on Elon Musk’s announcements: in typical fashion, developments come quickly (e.g., Grok 5 could be around the corner, or xAI might open source Grok’s model as hinted ( ^[38] ), which would create another wave of derivative chatbots). The GEO lesson here is agility – those who adapt fastest to new AI platforms can capture early visibility before the space gets crowded.

8.4 Key Differences Among Major LLMs (GPT-4, Claude, Bard/Gemini, LLaMA, Grok)

The emerging landscape of LLMs is not homogeneous – each model comes with its own architecture, training data, update cycle, and use case focus. These differences have strategic implications for how one approaches GEO. Let’s break down some key dimensions of variation among the leading models and why they matter: 1. Knowledge Freshness & Data Access: Perhaps the most immediately relevant difference is whether the model has access to current, real-time information or is limited to a fixed training cutoff. OpenAI’s ChatGPT (GPT-4) has a knowledge cutoff (September 2021 for the base GPT-4 model), meaning out-of-the-box it doesn’t “know” about events or content created after that date. To compensate, OpenAI introduced features like the Browsing mode (using Bing’s search) and Plugins that can fetch information or run computations. In late 2023, ChatGPT regained the ability to browse the web live when explicitly enabled, and by 2024 GPT-4 could handle limited live queries via Bing integration. However, most casual ChatGPT interactions still rely on the static knowledge unless the user actively invokes a plugin or browsing. This means if your content is newly published, there’s a lag before GPT-4 knows it by default. It might appear in GPT’s answers only if a user’s query triggers the browsing mode or if OpenAI has done a new training data refresh (OpenAI does periodically fine-tune models with more recent data or user-provided data, but these updates are infrequent). For GEO, this indicates that timely content (news, trends) might not surface via ChatGPT unless you ensure it’s also findable through search (so that a browsing GPT-4 can retrieve it) or unless it’s included in a popular plugin’s data source. Google’s Bard / Gemini is at the other end of the spectrum – it was built with live data access in mind . Bard is connected to Google Search in real time. Every query to Bard can pull fresh results from the web (much like a search engine would), and Bard will incorporate those into its answers. Google’s next-gen model Gemini (which Bard is transitioning to) continues this live data approach, and internal reports suggest it can handle even larger context (there are claims of Gemini models with up to 1 million tokens context window when including retrieved data) ( ^[39] ) ( ^[40] ). Practically, if you publish a webpage and it gets indexed by Google, Bard can start quoting it within hours or days . We’ve seen Bard give answers with citations linking directly to a freshly updated website or a breaking news article. So, for real-time GEO , Google’s AI demands that you maintain extremely up-to-date content . If you have product pricing, for example, and Bard is sourcing answers about pricing, an outdated price on your site could propagate immediately into Bard’s answer box. The lesson is to sync your content with reality as much as possible – treat it like how you’d approach voice search or featured snippets, which also prioritize current and accurate info. Moreover, since Bard integrates deeply with the Google ecosystem (e.g., it can pull data from Google Maps, Google Travel, etc.), if local SEO or any Google-specific feature matters to you, it matters for Bard as well. For instance, Bard might answer a local query by drawing on Google Business Profile info or Google reviews. Anthropic’s Claude , as discussed, doesn’t have built-in web access and relies on its training data (up to 2022 or so) plus any user-provided documents/links. In practice, this makes Claude similar to ChatGPT’s base mode for external info – it won’t know about recent developments unless you feed them to it. However, in integrated contexts like Slack, a user can share a link for Claude to read ( ^[41] ). So optimization for Claude involves making sure that if someone were to feed your content to an AI, it’s ready to be consumed (clear language, well-structured, no login barriers). Also, Anthropic tends to update Claude’s model less frequently than Google updates Bard. As of late 2024, Claude 2 was the main public model, and Claude 4 (internal name for an upgraded version) came in mid-2025 – these models would have training data that might include snapshots of the web up to certain dates (Anthropic hasn’t publicly detailed cutoffs, but a Slack FAQ implies a ~2-year lag ( ^[7] )). So think of Claude’s brain as a knowledgeable person who hasn’t read the news in two years, but who can quickly read anything you hand them. For GEO, that means historical evergreen content is well-represented in Claude (if your site had high-quality pages in 2020–2022, Claude likely “knows” them), whereas brand new content is invisible unless actively provided. Meta’s LLaMA (and other open models) vary. If someone is using an off-the-shelf LLaMA 2, its knowledge reflects its 2023 training data. If they fine-tuned it on domain-specific corpus, it knows that corpus up to whatever cut. Open models used in specialized apps may not have any live update mechanism (unless the app builds one via retrieval). However, because anyone can fine-tune, we do see creative approaches: some open-source projects do weekly or monthly fine-tunes on the latest Wikipedia or StackExchange dumps, for example, to keep a model semi-fresh. Still, none of the open models (as of 2025) have the robust live crawling that Google Bard does out of the box. So for open models, the key is getting into their training or retrieval sets . Many open LLMs use Common Crawl or RedPajama (an open dataset) – ensuring your site isn’t blocking those crawlers and maybe even contributing to open data (like Wikipedia) helps. One advantage: if an open model is used via a tool like Perplexity AI , that tool does live crawl and then feeds the text into the model with a prompt. In that scenario, it behaves a bit like Bard – your fresh content can be fetched on the fly. It highlights that the interface or application layer matters : an open model plugged into a search engine will have current knowledge (via retrieval), whereas the same model in offline mode will not. GEO strategy should thus account for both possibilities. 2. Context Window (How Much Content the AI Can Handle at Once): We touched on this with Claude’s 100K tokens, but let’s compare: Claude 2/Claude 4 : ~100K to 200K tokens (the largest in industry). This is a huge deal for tasks like analyzing long texts, as mentioned. For GEO, this means Claude can effectively read your entire website section or a full PDF. If someone asks Claude, “Summarize Acme Corp’s 2022 Sustainability Report,” Claude could take the whole report (if provided) and do it. For marketers, if you produce large reports or extensive documentation that you hope AI will accurately represent, Claude is your friend in the sense that it won’t easily lose pieces of context. However, note one limitation: just because it can read 100k tokens doesn’t guarantee perfect summarization. Anthropic themselves note Claude might still hallucinate details not present ( ^[42] ) or make errors if asked to juggle too many instructions ( ^[42] ). But overall, it’s best-in-class for long inputs. OpenAI’s GPT-4 : Initially launched with an 8K token window, and a 32K version for limited users. By late 2024, OpenAI announced GPT-4 Turbo with up to 128K tokens for developers (though the ChatGPT interface for most users was still limited to 8K or 32K) ( ^[39] ) ( ^[43] ). A 128K token context begins to rival Claude’s; it’s about 96,000 words. This was a significant upgrade likely to keep pace with Claude. It means some enterprise or advanced users of GPT-4 could feed nearly a book-length text into it. For GEO, it implies that lengthy content optimization isn’t about splitting into many small pages for AI – an AI can take the whole thing. Instead, focus on structuring long content (using clear headings, summaries) so that when an AI with a big context reads it, it can identify the important parts. Also, if you expect GPT-4 (through Bing Chat perhaps) to read your content, you can make their job easier by including an executive summary or conclusion section that succinctly wraps up the piece. The AI might latch onto that for its answer. Google Gemini (Bard) : According to some reports, Gemini’s advanced versions push context limits even further – experiments with 1 million tokens ( ^[44] ) and talk of even 2 million in the future ( ^[40] ). Those numbers are staggering (that’s basically an entire library of documents in one go). It’s unclear if those will be broadly available or just research demos. But suffice it to say, Google is aiming for the AI to effectively “read the whole internet if needed” for a query. Already, Bard’s integrated search results might include say 3–5 web pages of information combined. If Gemini can really scale to hundreds of thousands of tokens, it might digest entire topic clusters at once. The takeaway: depth of content will be fully accessible. So old SEO wisdom of splitting content into multiple pages for higher page views is moot for AI – one strong, comprehensive page on a topic is better, because an AI can consume it all and there’s no notion of click fatigue or bounce rate with an AI reader. Moreover, if your competitor has the most definitive 50-page guide on a subject, an AI like Gemini could effectively extract the key points from all 50 pages and you won’t beat it by just having a 1-page shallow article. In fact, the AI might never surface your shallow article if it finds the comprehensive source first. This argues for covering topics comprehensively and in one place (or at least interlinking them well) so that an AI sees your content as a one-stop resource. LLaMA and other open models : Out of the box, many open models have modest context (e.g., 4K or 8K tokens). But the community has developed techniques to extend context (like position interpolation, etc.), and some fine-tuned versions support 16K or more. Still, they generally lag behind Claude/GPT-4 in this department. That means many community-run bots can’t handle super long inputs. If you’re optimizing for something like a local instance of LLaMA that a user might query (maybe via a mobile app), you might still want to structure content in chunks. However, open models will likely catch up; it’s just a technical arms race. One interesting facet: since open models can be combined with vector databases for retrieval, a user could ask a question and the system fetches multiple relevant snippets from your content and feeds them in. In that case, the effective context could be large (split across many retrieved chunks). To ensure your content is retrieved, be sure to use clear keyword-rich headings and paragraph summaries – the retrieval algorithms (vector similarity) often depend on semantic relevance, so covering subtopics in a distinct way can help your content be selected as a relevant chunk for an answer. 3. Multimodality (beyond text): Some LLMs handle images, audio, and more, while others are text-only. GPT-4 (Vision) : OpenAI introduced a version of GPT-4 that can accept images as input. This means ChatGPT (for some users) can analyze an image you upload – describing it, interpreting charts, reading screenshots (OCR), etc. From a GEO perspective, this means that visual content on your site could be parsed by AI . If an AI is browsing your page and there’s an infographic, GPT-4 might actually read the text from that image or describe the chart to answer a question ( ^[39] ). Therefore, it’s important to include alt text and captions for images as usual (good for accessibility and now for AI understanding). Also, providing transcripts for videos or text for diagrams ensures no information is lost to an AI trying to be helpful. In the future, we might see AI answers that say, “Here is a diagram from Brand X’s report [diagram interpretation]” if allowed. Already, Bing Chat (which uses GPT-4) will sometimes answer questions about a chart by actually analyzing the image content if the URL is provided. Marketers should assume anything visual may be extracted by AI – so if there’s key data in an image, also put it in text on the page. Google Gemini is rumored to be multimodal from the ground up (DeepMind’s expertise in images and video is being folded in). We can expect Bard to eventually handle images or even video frames. Google has already shown demos of their models summarizing YouTube videos or answering questions about an image (Google Lens + Bard integration). So think about visual GEO : e.g., if someone asks the AI about “what does the new Acme product look like?”, a multimodal AI might actually pull an image of it (maybe from your website or a user’s social media) and then describe it or show it. Ensuring that your official images are high-quality, easily discoverable (proper SEO for images), and accurately represent your products could influence what the AI presents. Also, consider watermarking images – either visually or via metadata – so if they are used by an AI, your brand is subtly in there. Google has talked about watermarking AI-generated images; on the flip side, you might want to watermark real images so AI doesn’t accidentally attribute them incorrectly. Other models : Claude is currently text-only (though Anthropic might experiment with multimodal in the future). Grok, interestingly, introduced “Eve”, a voice that can speak answers in a British accent ( ^[45] ), showing xAI’s interest in audio output . It didn’t mention image input yet, but voice output means potentially voice input (especially in Teslas). If voice search 2.0 becomes voice conversation with AIs like Grok, the phrasing of your content (conversational tone, easy-to-read-aloud text) will matter more. In Chapter 14 we discuss voice search revival, but it’s worth noting here as models differentiate: some will be the voice assistants of the AI era (e.g., integrated in cars, phones), some will be the text analysts (like Claude in Slack). Tailoring content style to each (for instance, being succinct and clear for voice responses, vs. deeply informative for text analysis) might be a future split in GEO tactics. 4. Integration and Ecosystem: Each model lives in certain platforms: GPT-4/ChatGPT : Available via OpenAI’s ChatGPT interface (web and mobile app), via API for businesses, and notably integrated into Microsoft’s ecosystem . Microsoft’s Bing Chat uses GPT-4 (with search) and is accessible in Edge browser, Windows 11 (the Copilot sidebar), and even Office apps for writing assistance. So if your target user is, say, using Word and asks the built-in AI to draft something about your industry, GPT-4 will be doing that under the hood. Also, ChatGPT plugins form a mini-ecosystem – companies like Kayak, OpenTable, and many others built plugins to allow ChatGPT to query their data specifically. As a marketer, you might consider creating a ChatGPT plugin for your service (if you have a lot of data or functionality to offer). For example, a travel company could have a plugin so when users plan a trip in ChatGPT, the bot can pull live pricing from the company’s system. If relevant to content, having a plugin that feeds your latest blog posts or product info into ChatGPT could ensure the AI always has the freshest from you without needing general web access. There’s a barrier to entry (technical and approval by OpenAI), but it’s an avenue to directly embed your content into GPT-4’s world . Google Bard/Gemini : Obviously part of Google’s search and services. Bard is gradually being tied into Chrome (with features like “Google it” and seeing the source of Bard’s info in search results) and likely will be part of Android (there are reports of Bard integration with Google Assistant). If you think of traditional SEO, it was all about the Google ecosystem (search, featured snippets, etc.). Now with Bard, it’s still Google’s world but with generative answers. The AI overviews in Search Generative Experience (SGE) are an example: Google will synthesize an answer to a query at the top of search, often with citations or links to sources ( ^[46] ). Ensuring you rank or at least get cited in those overviews is classic SEO + some GEO (which in Google’s case, means all the things you did for SEO – quality content, schema, etc. – remain crucial because they help the AI identify trustworthy info). In Chapter 6, we detailed optimization for Google’s AI, but suffice to say, if your content adheres to Google’s EEAT and technical SEO guidelines , you’re indirectly optimizing for Gemini’s outputs as well. Claude (Anthropic) : Mostly accessed via API (integrations like Slack, or platforms like Notion, or on AWS Bedrock), and via Anthropic’s own interface (Claude.ai). It’s also on Poe (Quora’s app). This means Claude is often behind the scenes, not a direct destination like ChatGPT or Google. Users might not even know an app uses Claude – they just know the app has “AI features”. For GEO, you might not specifically target “Claude” as a channel but think of the apps that use Claude . For example, if Slack’s AI (Claude) is big in workplaces, maybe publishing content that is useful in workplace contexts (like a well-researched whitepaper that someone might share with the Slack bot to summarize) could be a subtle strategy. Or if an enterprise platform uses Claude to power a customer support chatbot, feeding it good data via your knowledge base (assuming the chatbot indexes your KB) is key. Essentially, with Claude and similar, optimize for the end-use application rather than the model . Know where Claude is popular (e.g., lots of devs might use Claude via API for code assistance because of its large context). If you run a developer website, note that Claude might be reading large chunks of documentation to assist programmers. So ensure your API docs or tutorials are clear, because an AI might be summarizing them to answer a developer’s question like “How do I implement X using Y API?”. Meta’s LLaMA : As an open model, it doesn’t have a single user-facing product. But interestingly, Meta might integrate it into their own products (rumors of AI features in Facebook, Instagram). Also, a version called Llama-2-Chat was accessible on platforms like Hugging Face and through Microsoft’s Bing (in a limited way, Microsoft experimented with an open model for some Bing chat users). For GEO, not much direct action is needed except to be aware that many smaller apps might be running on LLaMA . If you see a new app or website boasting an “AI assistant,” and it’s not saying it uses OpenAI or Google, chances are it’s using LLaMA or another open model under the hood. The quality of the answer in those apps will depend on how they’ve tuned that model and what data they’ve fed it. If it’s important, you might reach out to those app developers and ensure your content is being used properly (some companies might even offer an API or data feed for third-party AI apps – e.g., a medical association might license their content to a healthcare AI startup using open models, to ensure accuracy). xAI’s Grok : Integrated with X (Twitter) and soon Tesla, as mentioned. That means it’s partly a social media feature and partly a physical device feature (cars). For social media, one interesting tactic: since Grok can be “summoned” by tagging its account on X ( ^[47] ), brands or community managers could engage with it publicly. For instance, someone asks @Grok a question about your product on X; if Grok answers and it’s wrong, your brand account could reply with a correction. Not only do human users see that, but Grok might see your correction tweet (since it monitors X) and actually learn from it or use it in future answers. It’s a weird dynamic – effectively giving feedback to an AI through the public social channel. Musk’s vision is likely users casually interacting with Grok like another user on X. So treat Grok somewhat like an influencer or user: ensure it “hears” the right info. That could mean posting content on X that’s aimed at informing Grok (and people). For Tesla integration: that veers into voice assistant territory (like Siri/Alexa). If people start asking their car AI about local businesses or products while driving, make sure your local SEO is strong (since it might use location data + its model to answer). For example, “Hey Grok, where can I get a good coffee around here?” might trigger a response based on real-time Google Maps or X data (depending on what Grok taps into). While that’s speculative, it shows how cross-domain GEO can get – you may need to optimize traditional local search info because an AI in a car might use it. 5. Output Style and Tendencies: Each model has a “personality” in how it frames answers: GPT-4/ChatGPT tends to give pretty verbose, well-structured answers with a neutral tone. It often provides step-by-step explanations or bullet points if appropriate. It’s quite creative and eloquent , which is great for users looking for depth, but sometimes it means shorter factual answers get a lot of padding. For GEO, this means if you want ChatGPT to present your info, you need to ensure the facts are there (it will add the fluff on its own). Also, ChatGPT is trained to cite less unless specifically asked or using a plugin (because its default mode is not connected to live sources). So it might mention your brand without a link, just from memory. It might say “According to [YourBrand]’s blog, doing X is beneficial” if it recalls that, but it won’t give the URL. Interestingly, users often copy-paste ChatGPT answers into search if they want the source. Thus, a user might see something that sounds like it came from your site and then search it – hopefully finding your site. To facilitate that, using distinct phrasing or branded keywords in your content can create that connection. E.g., if you coin a term or have a unique slogan, ChatGPT might reproduce it verbatim (since it was in training data), and then a user could trace it back to you. We saw this with things like recipes or poems – ChatGPT would regurgitate a specific blogger’s recipe text, and people could find the source via Google. That’s not ideal (it’s essentially plagiarism by the AI), but it happens. Watermarking content with subtle signatures is a debated strategy – OpenAI was working on watermarking AI outputs, but here we’re talking watermarking inputs for recognition. No clear solution there yet, beyond being aware of it. Claude has a neutral, friendly tone and tends to be extremely verbose (sometimes overly so). It often restates the question and goes into exhaustive detail. Anthropic geared it to be helpful and transparent; it sometimes even disclaims its limitations (e.g., “I’m an AI, I don’t have feelings but you asked about…”) more than others. For GEO, if a user is getting an answer via Claude, they’re likely getting a thorough exposition. If you want a specific snippet of your content to shine in Claude’s answer, it helps if that snippet is clearly the most relevant piece. Claude might include multiple viewpoints or a pros/cons list. So if, say, you have an article about the pros and cons of a product, making that list explicitly will increase the chance Claude picks up that structure and uses it. Claude also has the concept of a “constitution” guiding it, which includes things like not being misleading, not giving harmful advice, etc. If your content is borderline (e.g., financial advice, medical claims), Claude might handle it more carefully or even refuse details compared to a more lax model. So ensure such content on your site is well-balanced and not extreme, or the AI might flag it as not fully trustworthy to repeat. Bard/Gemini initially tended to be more concise and factual (sometimes to a fault, e.g., a bit dry). Google has tuned it to improve creativity over time, but many users found Bard’s answers shorter than ChatGPT’s. Bard is also more likely to point to sources (with the little “Google it” button or inline citations in SGE). One advantage: if Bard cites you, it’s giving a hyperlink directly to your site. To earn that citation, your content likely needs to match the query intent closely and be authoritative. Google’s algorithms for SGE citation aren’t public, but presumably they consider the same signals as for featured snippets: relevance, authority, and maybe the presence of exactly the text that the AI output. If Bard’s output says, “According to Acme.com, 90% of customers prefer personalized ads 【source】,” it likely grabbed that stat from a line on Acme.com. Providing such succinct, quotable stats or statements in your content can make it easier for the AI to just lift that and attribute you. In contrast, if your content is all narrative and the stat is buried in a paragraph, the AI might summarize it without quoting, leading to a generic answer. So consider structuring key facts in standalone sentences or bullet points that are ready for citation . Also, Bard being part of Google means it respects certain things like schema markup . For example, if you use FAQ schema, Google might use those FAQs in Bard’s training data or to directly answer questions (much as it does for featured snippet answers). Continuing to implement structured data (like HowTo, FAQs, Product schema) is beneficial – if Google AI can easily parse your content’s meaning, it might prefer it to some unstructured text. Grok we discussed as having a witty, sometimes irreverent tone . It might produce answers with memes, sarcasm, or casual language. If a user base prefers that style, they might gravitate to Grok for entertainment as much as answers. For brands, this means if Grok picks up on a joke about your brand, it might propagate it. Embrace humor as part of your content strategy if it fits – e.g., having a sense of humor on social media means if Grok shares an answer involving your brand, it might use your own humorous post (which is better than a nasty joke someone else made). On the flip side, correct false humorous takes – if there’s a running joke based on incorrect info, publicly clarify it, so Grok gets the memo through the X data. We can summarize some of these differences in a quick comparison table for clarity: ( ^[46] ) ( ^[39] ) LLM (Provider) Data Freshness Max Context Notable Strengths Notable Limitations GPT-4 (ChatGPT, OpenAI) Training cutoff 2021; optional Bing browsing & plugins for updates8K tokens (base); 32K extended; up to ~128K in beta ( ^[39] ) ( ^[43] )Highly fluent & creative; broad knowledge; plugin ecosystem; large user base (~122M daily ( ^[48] )); multi-modal (image inputs)Static knowledge without browsing; may fabricate sources; strict content filters reduce some answers (avoids controversies) Google Bard (Gemini) Live Google Search results for every query (real-time info)Very high (rumored 1M+ tokens in experimental versions ( ^[49] )); effectively unlimited with retrievalUp-to-date information; deeply integrated with Google services (230+ countries, 40+ languages ( ^[50] )); cites sources in SGE; strong logical reasoning (Gemini improved chain-of-thought)Still improving creativity; smaller market share so far (~13% of chatbot use ( ^[51] )); answers can be terse; reliant on correct search index data Anthropic Claude Training data ~2022; no built-in web access (reads links on request)~100K tokens (Claude 2) up to 200K (Claude 4) ( ^[3] )Massive context (reads long documents); very detailed and structured responses; “Constitutional AI” yields safe, transparent answers; popular in enterprise integrations (Slack, etc.) ( ^[46] )Limited knowledge of recent events; small direct user base (~2–3% of chatbot market ( ^[52] ) ( ^[53] )); can be verbose and sometimes overly cautious (may refuse speculative queries) Meta LLaMA 2/3 (Open-Source) Varies by implementation (usually static data; can be fine-tuned on new info; many rely on user-provided retrieval)Typically 4K–16K tokens (community extending this gradually) Customizability – can be fine-tuned for any domain or language; no licensing fees; large community creating variants; powering countless niche apps (education, medicine, etc.) ( ^[17] ) ( ^[18] )Quality depends on fine-tune (can be inferior to GPT-4 without domain data); no single trusted interface (consistency issues); usually no direct internet access unless added by app developer; smaller contexts unless specialized architecture xAI Grok Training on public web data and real-time X/Twitter firehose ( ^[28] ); up-to-the-minute social content(Not publicly stated; likely tens of thousands of tokens, but smaller than Claude/Gemini)Live social media insight (trends, memes); bold and entertaining tone; fast iteration (4 model versions in ~1 year ( ^[54] )); will be embedded in X platform and Tesla (unique distribution)Unpredictable outputs (less filtering – risk of offensive or incorrect content ( ^[55] )); not yet widely used (niche audience); factual accuracy can be uneven if social data is noisy; currently requires X Premium for access (limited pool) ( ^[56] ) – that’s roughly 1 in 8 people on Earth. Of those, about 60% of usage share is ChatGPT, ~14% Bing Chat, ~13% Bard, ~3% Claude, and the rest others (Perplexity, local assistants, etc.) ( ^[57] ) ( ^[37] ). Regionally, it varies: in China, Baidu’s ERNIE Bot reportedly hit 300 million users shortly after launch ( ^[58] ) (due to a huge domestic user base); in the West, ChatGPT reached 100 million users in just 2 months – the fastest adoption ever ( ^[59] ) – and by early 2025 had ~800 million weekly users ( ^[48] ). The point is, people are flocking to these new tools , but not all to the same one. Where once “Google = search” for most, now the traffic is split among platforms. Some estimates put ChatGPT at 5th most visited site globally by 2025 ( ^[60] ), but Google.com is still up there too. And Bing, thanks to AI, saw a 40% jump in usage to ~140 million daily users by 2024 ( ^[61] ), carving out a measurable chunk of search market. So how do we adapt to this multi-model world? Here’s a roadmap of strategies: A. Maintain and Strengthen Core SEO : Your foundation is still your website content and its traditional SEO. Why? Because almost all AI systems ultimately rely on web content either in training or in retrieval. Ensuring your site is crawlable, fast, mobile-friendly, and indexed remains crucial. For example, Bing’s AI won’t surface your info if Bing’s crawler can’t index your site well. Google’s SGE won’t cite you if you’re buried on page 5 of results or if your content quality is low. And many LLMs (including open ones) ingested their knowledge from Common Crawl or Google results of the past. So the old advice of “optimize for humans, with search engines in mind” now extends to “optimize for humans, with search engines and AI in mind.” Luckily, the practices overlap: clear site structure, good metadata, schema, authoritative backlinks – these all help AI find and trust your content. **In the AI age, “SEO content” is not some keyword-stuffed text for ranking; it is your brand’s knowledge, likely to be directly consumed by AI with no intermediary. So invest in quality. B. Embrace Structured Data and Feeds : Different AI agents might consume content in different formats. Some might prefer a nicely formatted HTML page; others might pull from a JSON API if available. Providing structured data (like schema.org markup for products, FAQs, how-tos, reviews, etc.) makes it easier for AI to identify key facts about your brand or offerings. Google’s AI, for instance, could use schema to provide concise info (much like featured snippets or knowledge panels do). If you have an e-commerce site, using product schema means an AI can confidently give details like price, availability, and ratings in an answer. Additionally, consider offering an API or data feed for your content. We’re already seeing this with some publishers creating APIs for their content so that AI companies can use them (sometimes via partnership deals). While a small business might not negotiate with OpenAI, you can still make a public data feed – for example, an RSS feed of blog posts (some AI like Bing’s index may check RSS for updates), a sitemap that’s always updated (for crawlers), or even a dedicated “AI endpoint” if you’re tech-savvy that returns info in a structured way. One interesting idea: some brands are creating ChatGPT plugins (as mentioned earlier) which is essentially offering an API to OpenAI’s ecosystem. If users install your plugin, ChatGPT will fetch real-time data from you when relevant. Plugins are a form of structured access to your content or services. C. Distribute Content Across Key Platforms : To catch users on each AI, you may need to go to where the AI is . For ChatGPT, that could mean having a presence on their plugin store or writing content on forums like OpenAI’s community (where prompts and best answers get shared – indirectly influencing the AI). For Google’s ecosystem, it could be continuing to nurture your YouTube channel or Google Business Profile (since those can feed into Google’s AI answers for multimedia or local queries). For Bing, leveraging Microsoft’s properties like LinkedIn (Microsoft owns it) might indirectly help – e.g., Bing’s AI can see LinkedIn Insights articles for B2B queries. For open-source AI, consider contributing to open data sources: if you’re in academia, publish in arXiv (LLMs ingested lots of arXiv papers); if you have how-to guides, contribute to WikiHow or similar which are scraped into datasets; if you have definitions or knowledge, add them to Wikipedia or niche wikis. Essentially, content syndication – in a tailored way – ensures you’re present no matter which AI knowledge base is consulted. A trivial example: define your brand on Wikipedia. Many LLMs will answer “What is Brand X?” with whatever Wikipedia says (if it has an entry). If you don’t have one, the AI might struggle or use random data. D. Leverage International and Local AIs : If your market is global, you can’t ignore regional AIs. China’s Baidu ERNIE, Tencent’s Hunyuan, or Alibaba’s Qwen are the go-to’s in China (since ChatGPT is blocked there). They have their own ecosystem: e.g., Baidu’s Search with AI results, or WeChat bots. Ensure your content (translated appropriately) is accessible on Chinese platforms – for instance, if you have Chinese customers, having a Baidu Zhidao (Q&A) presence or posting on WeChat public accounts could get your info into those models. Similarly, South Korea might have its own (Naver’s HyperCLOVA), or the Middle East (there’s a surge in Arabic LLMs). While you may not actively optimize for each, at least be aware and ensure no technical blocks. For example, if you blocked all non-Western crawlers via robots.txt or Cloudflare thinking they were irrelevant, you might be invisible to an entire region’s AI systems. Open up where feasible. E. Monitor AI Mentions and Context : Traditional SEO has us monitor rankings and traffic. Now, we also need to monitor AI outputs . This is tricky because you can’t scrape ChatGPT or Bard easily (and their answers vary by prompt). But you can crowdsource by asking your community or employees to test certain key queries on different AI platforms and report back. For instance, if you’re a hotel chain, ask: “Plan a trip to Paris” on all the AIs and see if your hotels get mentioned in the recommendations. If not, what sources are they citing or pulling from? Maybe TripAdvisor or some blog – that indicates those sources hold sway in AI answers, so you might need to ensure your info (and correct info) is on those platforms. There are also emerging tools (AI visibility trackers) that claim to analyze AI results and see if your brand appears. While early, consider using them for critical topics. If you find misinformation (e.g., an AI says your product has a defect which is untrue), take steps: correct it on your site, publish a clarification press release, and maybe even use the AI’s feedback channels . OpenAI and others have feedback forms where you can report erroneous information. They do use this to improve models. For example, if ChatGPT falsely says “Product X was recalled in 2022” and you report it with evidence, OpenAI might adjust the fine-tuning to fix that. It’s akin to online reputation management but with an AI twist. F. Adapt Conversion and Measurement Strategies : In a multi-model world, a user might get all the info they need from the AI and never visit your site – yet still decide to buy your product or use your service. The classic web traffic metrics may not fully capture your reach. For instance, if ChatGPT recommends your product and the user later goes directly to your Amazon listing to purchase, your web analytics see nothing, but the AI essentially drove the sale. This calls for new ways to measure AI-driven referrals . Some ideas include: adding survey questions “How did you hear about us?” with an option for AI assistant; tracking aggregate trends (e.g., do sales correlate with spikes in certain AI queries trending?); or using unique identifiers in content so if they do copy from AI to site, you can catch it. It’s early days, but marketers are looking at AI visibility metrics . Some tools try to estimate “share of voice” in AI answers. You might not get precise numbers, but qualitatively keep an eye on this. Also, consider the role of branding : in an answer with no links, just text, having your brand name mentioned is crucial. If an AI says “a leading brand offers this solution...” that’s a lost opportunity if it doesn’t name you. Encourage the use of your brand name in content pieces and third-party articles (so that AI picking up that info includes the name). Building a strong brand will matter even more – users might start preferring to prompt AIs with “Using info from [Brand]’s site, give me XYZ” if they trust you, or they might double-check AI answers by specifically asking about your brand’s perspective. G. Prepare for Multi-Modal and Multi-Turn Engagements : Users might engage with an AI agent through an entire customer journey – asking initial questions, drilling down, then finalizing a decision – all without a search engine or website. To not be sidelined, you may need to participate in those multi-turn dialogues . How? Potentially through AI agents or plugins that insert at the right time. For example, OpenAI envisions tools that can proactively offer help in a conversation. If a user is talking with ChatGPT about financial planning for 5 turns, maybe a plugin for a finance calculator or a robo-advisor pops in (with user permission). If you have such a tool or informational widget, aligning with those frameworks could put you directly in the conversation. Another approach is to supply content formatted as Q&A or conversational snippets (some companies create conversational FAQs anticipating how AI might use them). Think in terms of dialogue : what follow-up questions does your content raise, and have you answered them somewhere? If not, consider expanding content to fill those gaps, so the AI doesn’t have to fill them (possibly incorrectly). H. Focus on Trust and Accuracy : In a sea of AI answers, users will gravitate towards those they trust. If your brand is consistently cited or referenced as a reliable source, that builds credibility. Conversely, if an AI often delivers wrong info about you (or none at all), that erodes trust. Work on digital PR : make sure authoritative sites (news, industry orgs) carry correct information about you, because AIs trained on that corpora will take it as ground truth. Also, content that demonstrates your Experience and Expertise (the first two E’s of EEAT) – like case studies, original research, expert quotes – can make AI answers incorporate that richness, essentially differentiating the responses when your info is used. For example, if your article includes a quote from your CEO with a unique insight, an AI might include that quote in an answer, which lends a human touch and authority to the response. Ensuring the AI doesn’t hallucinate is partly out of your control, but you can make it easier for the AI to be accurate by publishing clear facts and correcting errors openly . In closing, navigating the multi-model ecosystem means being everywhere your customers might query an AI . It’s akin to the early days of social media when companies realized they needed a presence on Facebook, Twitter, Instagram, etc., not just their own website. Now the “platforms” are AI assistants. It may sound overwhelming, but start with priorities: likely ChatGPT, Google/Bard, and Bing cover a large share in Western markets – so focus on those content and integration points first. Then expand to others as needed. The good news is that investing in high-quality content and data about your business pays dividends across all these AIs** – because they all ultimately rely on human-created knowledge. By making your brand’s knowledge broadly available, easy to consume by algorithms, and authoritative, you position yourself to be the answer no matter which AI is asked. The future will not have one monolithic search box but rather a tapestry of AI assistants embedded in our lives; by preparing now, you can ensure your visibility and influence remain strong as this transformation unfolds.

References

Anthropic News – Introducing 100K Context Windows (May 11, 2023) ( ^[1] ) ( ^[2] ) Anthropic Documentation – Claude in Slack FAQs ( ^[7] ) ( ^[8] ) Google Cloud Case Study – Quora’s Poe powered by Claude ( ^[9] ) ( ^[10] ) Meta News – How Companies Are Using Meta Llama (May 7, 2024) ( ^[21] ) ( ^[22] ) Meta Press – Meta and Microsoft Introduce Llama 2 (July 18, 2023) ( ^[16] ) ( ^[13] ) Reuters – Baidu’s ERNIE Bot hits 300M users (June 28, 2024) ( ^[58] ) Business Insider – What is Grok? xAI’s Chatbot (July 10, 2025) ( ^[27] ) ( ^[30] ) xAI Blog – Grok 3 Beta: The Age of Reasoning Agents (Feb 19, 2025) ( ^[62] ) ( ^[63] ) The AI Enterprise – Musk’s Grok – Irreverent Chatbot (Feb 3, 2024) ( ^[28] ) ( ^[29] ) DataStudios Report – ChatGPT vs. Google Gemini vs. Claude (June 12, 2025) ( ^[3] ) DataStudios Report – Most Used AI Chatbots in 2025 (May 25, 2025) ( ^[48] ) ( ^[51] ) StatCounter – AI Chatbot Market Share, Worldwide (June 2025) ( ^[37] ) Reuters – Grok AI in Tesla vehicles (July 10, 2025) ( ^[34] ) ( ^[55] ) Wikipedia – Grok (xAI) Logo description ( ^[64] .png (Used for reference on Grok branding, public domain logo).

References

^[1] www.anthropic.com - Anthropic URL: https://www.anthropic.com/news/100k-context-windows

^[2] www.anthropic.com - Anthropic URL: https://www.anthropic.com/news/100k-context-windows

^[3] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[4] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[5] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[6] www.anthropic.com - Anthropic URL: https://www.anthropic.com/claude-in-slack

^[7] www.anthropic.com - Anthropic URL: https://www.anthropic.com/claude-in-slack

^[8] www.anthropic.com - Anthropic URL: https://www.anthropic.com/claude-in-slack

^[9] Cloud.Google.Com Article - Cloud.Google.Com URL: https://cloud.google.com/customers/quora

^[10] Cloud.Google.Com Article - Cloud.Google.Com URL: https://cloud.google.com/customers/quora

^[11] Cloud.Google.Com Article - Cloud.Google.Com URL: https://cloud.google.com/customers/quora

^[12] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2023/07/llama-2

^[13] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2023/07/llama-2

^[14] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2023/07/llama-2

^[15] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2023/07/llama-2

^[16] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2023/07/llama-2

^[17] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[18] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[19] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[20] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[21] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[22] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[23] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[24] About.Fb.Com Article - About.Fb.Com URL: https://about.fb.com/news/2024/05/how-companies-are-using-meta-llama

^[25] Mistral.Ai Article - Mistral.Ai URL: https://mistral.ai/news/announcing-mistral-7b

^[26] Group.Ntt Article - Group.Ntt URL: https://group.ntt/en/magazine/blog/tsuzumi

^[27] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[28] www.theaienterprise.io - Theaienterprise.Io URL: https://www.theaienterprise.io/p/ai-rewind-elon-musks-ai-entry-grok

^[29] www.theaienterprise.io - Theaienterprise.Io URL: https://www.theaienterprise.io/p/ai-rewind-elon-musks-ai-entry-grok

^[30] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[31] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[32] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[33] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[34] www.reuters.com - Reuters URL: https://www.reuters.com/business/autos-transportation/grok-ai-be-available-tesla-vehicles-next-week-musk-says-2025-07-10

^[35] www.reuters.com - Reuters URL: https://www.reuters.com/business/autos-transportation/grok-ai-be-available-tesla-vehicles-next-week-musk-says-2025-07-10

^[36] www.socialmediatoday.com - Socialmediatoday.Com URL: https://www.socialmediatoday.com/news/x-formerly-twitter-rolls-back-changes-to-grok-ai-chatbot/752632

^[37] Gs.Statcounter.Com Article - Gs.Statcounter.Com URL: https://gs.statcounter.com/ai-chatbot-market-share

^[38] Voicebot.Ai Article - Voicebot.Ai URL: https://voicebot.ai/2024/03/12/elon-musk-will-open-source-grok-llm

^[39] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[40] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[41] www.anthropic.com - Anthropic URL: https://www.anthropic.com/claude-in-slack

^[42] www.anthropic.com - Anthropic URL: https://www.anthropic.com/claude-in-slack

^[43] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[44] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[45] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[46] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[47] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[48] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[49] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/chatgpt-vs-google-gemini-vs-anthropic-claude-comprehensive-comparison-report-capabilities-perfo

^[50] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[51] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[52] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[53] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[54] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[55] www.reuters.com - Reuters URL: https://www.reuters.com/business/autos-transportation/grok-ai-be-available-tesla-vehicles-next-week-musk-says-2025-07-10

^[56] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[57] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[58] www.reuters.com - Reuters URL: https://www.reuters.com/technology/artificial-intelligence/baidu-launches-upgraded-ai-model-says-user-base-hits-300-mln-2024-06-28

^[59] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[60] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[61] www.datastudios.org - Datastudios.Org URL: https://www.datastudios.org/post/the-most-used-ai-chatbots-in-2025-global-usage-trends-and-platform-comparisons-of-chatgpt-gemini

^[62] X.Ai Article - X.Ai URL: https://x.ai

^[63] X.Ai Article - X.Ai URL: https://x.ai

^[64] Commons.Wikimedia.Org Article - Commons.Wikimedia.Org URL: https://commons.wikimedia.org/wiki/File :Logo_Grok_AI_(2025

9. Content Strategy for GEO – Creating AI‑Optimized Content

In the era of Generative Engine Optimization (GEO) , crafting content requires a strategic blend of traditional SEO best practices and new techniques aimed at AI-driven search experiences . As search evolves from keyword-based queries to AI-generated answers, content marketers must adapt how they create and structure information online. This chapter explores how to produce content that not only ranks in classic search engines but is also favored and directly utilized by Large Language Model (LLM) systems like ChatGPT, Google’s Gemini-powered search, Bing Chat, Perplexity, Anthropic Claude, Meta’s LLaMA, xAI’s Grok, and other emerging AI tools. We will examine why demonstrating E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and originality is paramount, what types of content AI can’t easily replicate , how to format content for AI excerpting , the benefits of a conversational tone and FAQ-style organization, and the importance of continual content updates to remain relevant in the AI age. Throughout, we’ll include real-world examples (circa 2024/2025) and up-to-date statistics, and we’ll highlight tools and tactics for optimizing content to be discovered—and even quoted—by generative AI systems. By the end of this chapter, online marketing professionals will have a clear action plan for creating AI-optimized content that drives visibility and traffic in the generative search era.

9.1 E-E-A-T and Originality as Differentiators

One of the most critical factors for content success in generative search is adhering to Google’s E-E-A-T guidelines: Experience, Expertise, Authoritativeness, and Trustworthiness . High-quality, expert content matters more than ever because AI models tend to favor and surface information from sources that demonstrate credibility and depth ( ^[1] ) ( ^[2] ). In Google’s experimental Search Generative Experience (SGE) and “AI overviews,” for example, users noticed that well-known websites and brands (those with strong authority signals) were appearing prominently in the AI-generated answers ( ^[1] ). This suggests that Google’s algorithms, when selecting content to include in a generative snippet, lean heavily on perceived authority and trust – essentially, E-E-A-T still matters in the AI search era ( ^[3] ). Illustration of Google's E-E-A-T framework – Experience , Expertise , Authoritativeness , and Trustworthiness – as four pillars of content quality. In the AI-driven search landscape, content that demonstrates these qualities is more likely to be seen as reliable and thus favored by search engines and AI assistants. Ensuring your content showcases firsthand experience , expert knowledge ( expertise ), recognized authority , and solid trustworthiness can significantly improve its chances of being referenced in generative AI results. ( ^[4] ) ( ^[2] ) Why E-E-A-T is crucial for GEO: Generative AI tools like ChatGPT and Google’s upcoming Gemini model are trained on vast swaths of internet text. When these models formulate answers, they probabilistically draw on patterns learned from their training data. They don’t truly know which sources are authoritative, but their training process means that information echoed by many reputable sources – or by sources with strong digital footprints – carries more weight. Google’s own systems explicitly aim to surface high-quality info from reliable sources and to downplay content that lacks authority or trustworthiness ( ^[5] ) ( ^[2] ). In practice, this means that content written by recognized experts, content published on sites with a reputation for subject authority, and content that provides trustworthy, accurate information is far more likely to be selected by AI summarizers . As one industry guide noted in early 2024, Google is expected to “continue emphasizing E-E-A-T in content discovery results, whether in AI or regular search,” with the credibility of the page, author, and website all factoring into what gets displayed ( ^[6] ). In short, demonstrating E-E-A-T in your content isn’t just about pleasing human readers – it directly influences whether an AI will treat your content as a trustworthy source worth including in an answer. Originality and firsthand experience: A key aspect of E-E-A-T that **AI struggles to replicate is experience . Large language models can aggregate and rephrase common knowledge from the web, but they lack genuine firsthand experience or original insight . Content that showcases personal experience or unique expertise will stand out as something AI cannot simply generate on its own** ( ^[7] ). For example, a travel website that includes a blog post like “My 7-day trek through the Andes – lessons learned” (with vivid personal details, original photos, and first-person tips) is offering something qualitatively different from the generic travel summaries an AI might produce from Wikipedia and standard tourist info. That human touch – the experience – makes the content more trustworthy and valuable. In Google’s ranking guidelines, “experience” was added as a new facet to E-A-T in 2022, precisely to encourage content creators to share firsthand experiences when relevant (such as a product review written by someone who has actually used the product, or an analysis by a professional who has hands-on experience in the field). In the generative age, we can expect AI-driven search to similarly reward content that carries the stamp of personal experience and originality. Not only do human reviewers value this, but AI models trained on vast data can often detect when content is merely rehashing generic facts versus when it provides a novel perspective or real-life example. Indeed, marketers have found that AI-generated text often feels generic or “flat” without human insight – one marketing firm noted that “without a human perspective, AI-generated content can feel generic—or worse, misinformed” , underscoring that authentic human input is still critical ( ^[8] ). Authoritativeness and branding: Another facet of E-E-A-T is Authority , which in practice can relate to your brand’s reputation or your authors’ credentials. In the context of GEO, authority might be signaled by factors like: Are other sites referencing your content? Do your pages have quality backlinks? Is your brand well-known in the industry? Are your experts cited elsewhere? All these contribute to whether an AI might “think” of or prefer your content when constructing an answer. For instance, SEO professionals observed in 2024 that “big name” websites had an edge in Google’s AI overviews – likely because Google’s system associated those domains with authoritative content ( ^[3] ). This doesn’t mean smaller sites can’t get included, but it means establishing topical authority is key . Writing thorough, well-researched content and getting recognized (via references, mentions, or shares) by others in your field will help build that authority over time. Even on AI platforms like ChatGPT, which don’t link out by default, the underlying model is more likely to produce information it saw on authoritative sites during training. Thus, being cited on high-authority platforms (news sites, respected blogs, scholarly articles, etc.) indirectly influences LLMs. For example, if your company is frequently mentioned in industry reports or has a Wikipedia page , that information is in the training data of many models – increasing the odds that ChatGPT knows about it and might include it in relevant answers. In essence, digital authority translates to AI visibility , so investing in authoritative content (and promotion of that content through digital PR, thought leadership, etc.) is a strategic move. Trustworthiness and accuracy: AI models are notorious for sometimes generating confident-sounding but incorrect information (the so-called “hallucinations”). To minimize errors, LLM-based search experiences prefer content that is accurate and trustworthy . As content creators, this means double-down on fact-checking, citing reliable sources, and keeping information up-to-date. If your page contains claims or stats, reference the source or provide context – not only does this build trust with human readers, but those references might be part of what an AI uses to judge the veracity of your content. Google’s systems, for instance, place heavier emphasis on reliability signals for queries where accuracy is critical (health, finance, etc.) ( ^[5] ). For generative AI, if the model finds two conflicting pieces of information, it’s more likely to use the one that appeared in a context with other trust signals (for example, on a site with high E-E-A-T or in proximity to authoritative language). Practical tip: Showcase trustworthiness by having clear author bios (highlight credentials), including editorial policies or references, using HTTPS (security), and maintaining a clean site free of spammy ads or misleading layouts. All of these small factors can cumulatively enhance how your content is perceived by algorithms and users alike. Google’s stance on AI-generated vs human content: It’s worth noting that Google has clarified it doesn’t outright discriminate against AI-generated content – the key is quality . “Using AI doesn’t give content any special gains. It’s just content. If it is useful, helpful, original, and satisfies aspects of E-E-A-T, it might do well in Search,” Google stated plainly ( ^[2] ). In other words, Google cares about the end result, not whether a human or an AI wrote it. However, this also implies that **low-quality AI-written content that lacks originality or depth will not do well , and Google’s algorithms (and manual reviewers) are actively looking to down-rank or penalize content produced simply to game the system. In fact, in 2024 Google undertook major core updates to target spammy AI content. A notable example was Google’s April 2024 core update**, which “gave weighty punishments (including thousands of manual actions) to sites relying heavily on AI-generated content” that was of low value ( ^[9] ). Many websites that had flooded their pages with auto-generated text saw their search rankings plummet ( ^[9] ). The takeaway is clear: AI can be a useful tool in content creation, but it must be guided by human expertise and oversight . Mass-producing generic AI content is a recipe for disaster. Instead, if you use AI at all in writing, use it for initial drafts or outlines and then infuse human insights, originality, and editorial rigor into the final product ( ^[10] ). This aligns with the principle of E-E-A-T – the content should reflect human experience and trustworthiness, regardless of the tools used to create it. Actionable steps to boost E-E-A-T in content: Show your credentials: Attach author names to articles and include brief bios highlighting their expertise or experience in the topic. If your CEO writes a blog post on industry trends, mention their years of experience or notable achievements. Such signals help both users and AI gauge expertise ( ^[11] ). Cite sources and data: Whenever you present facts, statistics, or claims, back them up with references to credible sources. Not only does this build trust, but these citations could be picked up by AI models as part of the content’s context, reinforcing accuracy. Demonstrate experience: Where applicable, weave in personal experience or company-specific knowledge. For example, “At [Company], we tested this tool internally and found…” or “In my 10 years of practicing dietetics, I’ve observed…” . This kind of language can explicitly highlight experience (the new first “E”) to both readers and algorithms. Maintain a positive brand presence: Off-page factors contribute to authority. Encourage happy customers to leave reviews, participate in relevant forums or Q&As (like Quora, StackExchange), and collaborate with respected partners. All these build a web of trust around your brand name. As later chapters will detail (see Chapter 11 on Off-Page Signals), brand mentions across the web can influence AI outputs . For instance, LLMs like ChatGPT effectively treat a concept mentioned frequently across many contexts as more “true” or at least more salient ( ^[12] ) ( ^[12] ) . If your brand is consistently associated with a certain expertise (say your fintech blog is cited often regarding blockchain security), an AI is more likely to bring up your insights on that subject. Avoid manipulative tactics: Transparency is key to trust. Don’t hide sponsored content without disclosure, don’t stuff keywords or use cloaking. Google’s quality guidelines remain in force – any attempt to deceive users or the algorithm can backfire, especially now that AI systems might summarize exactly what’s on your page to users. You wouldn’t want an AI snippet to expose something like “This article doesn’t actually answer the question” or “The content appears auto-generated and thin” – which could happen if those elements are detectable. In summary, focusing on E-E-A-T and originality is about aligning your content with what both humans and AI find valuable : reliable knowledge, authentic perspective, and proven expertise. Generative AI is essentially an aggregator and amplifier of content patterns – if your content consistently embodies high quality and uniqueness, it increases the likelihood that AI will learn from it, select it, and propagate it when relevant user queries arise. By doubling down on quality and authenticity, you’re not just future-proofing for algorithms; you’re delivering real value to readers – which is exactly the point, after all.

9.2 Creating Content AI Can’t Easily Replicate

With the advent of advanced AI like GPT-4, Gemini, and open-source LLMs, it’s easier than ever to generate passable content on almost any topic. From generic how-to guides to basic product descriptions, AI can churn out “good enough” versions of widely available information in seconds . This raises a pivotal question for content strategists: What can we create that AI won’t simply generate itself? In other words, how do we make our content a “must-have” unique resource , rather than just another redundant web page? The answer lies in focusing on material that goes beyond easily scraped facts and common knowledge . Content that AI can’t easily replicate typically includes original research, proprietary data, deep analysis, strong opinions grounded in expertise, nuanced perspectives, and storytelling rooted in personal experience . These are the elements that make your content valuable and differentiating – not just to human readers, but also to AI systems that decide whether to quote or reference your site versus a hundred other sites saying the same thing. AI’s limitation – the lack of true originality: By design, generative AI models work by predicting likely sequences of words based on patterns in their training data. They excel at “regurgitating” the common denominator of what’s been published on a topic ( ^[13] ). For instance, ask an AI about the benefits of drinking green tea, and it will compile well-known points (antioxidants, improved focus, etc.) that appear across many articles. What AI does not do well is introduce completely new ideas or insights that haven’t already been extensively documented. It cannot truly originate a fresh research finding or recount a personal anecdote it never encountered. As an SEO expert succinctly put it, “AI is great at regurgitating common knowledge, but it struggles with original research, firsthand experience, and unique data.” ( ^[13] ). Google itself has acknowledged this in the context of search, emphasizing the value of “hidden gems” – content that provides unique insights not easily found elsewhere ( ^[13] ). For your content strategy, this means that if you produce the same cookie-cutter listicles or superficial content that dozens of other sites have, an AI overview will have no compelling reason to specifically include your phrasing or cite your page. Why would it, if you’re offering nothing new? On the flip side, if you publish something truly unique, then when users ask about that niche topic, your content is far more likely to be the one AI pulls in ( ^[14] ). Examples of content AI can’t easily replicate: Original research and data: If your company conducts a study or survey and publishes never-before-seen data, that’s gold. For example, imagine you run an email marketing platform and you analyze billions of emails to determine optimal send times or average open rates by industry. If you publish that report with detailed findings, AI models will eventually ingest that information as unique knowledge associated with your site. When someone asks, “What’s a good email open rate in retail?” an AI might actually cite or reference your data (especially if your study becomes frequently quoted by others, reinforcing its presence). In fact, marketers are increasingly doing this – Content Marketing Institute’s 2024 report noted a rise in brands creating data-driven content as a way to stand out, because it earns backlinks and trust. Unique statistics have always been link bait; now they’re also “AI bait.” Neil Patel’s team advises brands to “publish original research, case studies, or expert insights” because AI models favor content that offers fresh, data-driven perspectives that aren’t already ubiquitous elsewhere ( ^[11] ) . A real-world example: the site Ahrefs once did an original study showing that 90.63% of web pages get zero Google traffic (a striking statistic widely cited in SEO circles). That kind of original finding not only earned them human attention and backlinks, but any AI model trained on recent SEO content will also “know” that fact – and might mention Ahrefs or the stat in an answer about SEO challenges. The principle for you: invest in creating content that provides new data or insights – run a poll, analyze your user data (anonymously and ethically), perform an experiment or A/B test and share the results. This is content no one else has because only you had access to it. Case studies and in-depth analyses: AI can summarize generic best practices, but it cannot easily replicate a detailed case study of your client or your project. For instance, if you’re a marketing agency, writing a case study like “How We Increased Client X’s Conversion Rate by 50% in 3 Months – A Step-by-Step Breakdown” is highly valuable. It contains specific context, strategies applied, results, and lessons learned that aren’t published elsewhere. Another example: a cybersecurity blog might publish a teardown of a new malware strain based on the researchers’ hands-on analysis – again, unique content. These pieces serve a dual purpose: they are compelling to professionals in the field, and they provide fodder for AI that goes beyond the generic. If someone asks an AI, “How can e-commerce sites improve conversion?”, a generic model might list tips like “improve page speed” or “better CTAs.” But if your unique case study on improving conversion has been widely read and linked, a sophisticated AI (especially one like Bing or Perplexity that cites sources) might pull in a line from your case study, e.g., “One fashion retailer saw a 50% conversion lift by streamlining their checkout process ( ^[15] ).” The AI might even name the retailer or your blog if it was cited in training data or used as a source. By offering concrete, original examples , you give AI something specific to latch onto. Strong opinions and thought leadership: While AI can mimic a “bland consensus” of opinions found online, it tends to avoid taking controversial or very distinctive stances on its own. Content that includes a strong, well-argued opinion or a novel theory can thus stand out. For example, a technology analyst’s blog post titled “Why I Believe AI Chatbots Should Pay for Consuming Content” with a clear viewpoint and supporting arguments is unique content. If that perspective gains traction (perhaps it’s discussed on social media or other blogs), it might influence AI responses on the topic. A user might ask ChatGPT, “Should AI companies pay content creators?” and get an answer like, “There’s a debate on this. Some experts, such as [Name], argue that they should, citing reasons like X, while others say Y.” If your content is the one that started or exemplified that viewpoint, it may get a nod in the AI’s synthesized answer. Caveat: One must be careful that opinions are grounded in truth and not misinformation – LLMs also have a bias toward the majority view or well-established facts, and they try not to amplify fringe ideas unless prompted. But a well-reasoned expert opinion can become part of the “knowledge mix” for a topic , especially if it’s picked up by multiple sources. So don’t shy away from publishing insightful commentary or forecasts in your domain. It humanizes your brand and could become reference material. Storytelling and rich narratives: Storytelling – whether it’s customer success stories, personal journeys, or historical narratives – is another area where humans excel over AI. ChatGPT can produce a story, yes, but only based on patterns of stories it has read; it cannot replicate your story that has never been told. If you run a niche museum, an article like “The Lost Painting that Transformed Our Museum’s Fortunes – An Insider Story” will be one-of-a-kind. Humans love stories, and we remember them; AI, trained on what humans write and share, will “notice” a memorable story too. High-engagement content (measured by backlinks, time on page, social shares) is likely to bubble up in training data significance. So, if your story-driven content gains popularity, an AI might incorporate elements of it when relevant. Additionally, narrative content often includes many concrete details (dates, names, places) that could get picked up as facts by AI. For example, if your CEO writes “A Day in the Life” post describing how she uses your product, and it’s widely read, an AI might later answer a question about your product by saying “According to [Your CEO]’s account, she uses it every morning for XYZ.” Content with real E xperience: Expanding on the E-E-A-T “experience” point – content such as hands-on reviews, tutorials with step-by-step photos from real use, or field reports offer something AI cannot fake. A blog that shows “before and after” images of an actual home DIY project with personal notes will have unique value. Travel diaries, as mentioned earlier, or an entrepreneur’s first-person account of launching a startup – these are slices of reality. For instance, if you publish “Our Startup’s First 100 Days: What Went Wrong and What We Learned,” no AI can automatically produce that specific content because the experiences and mistakes are unique to you. If another founder asks an AI for advice on early-stage startups, the model might draw on insights from first-person accounts like yours (especially if such accounts become part of common knowledge in that sphere). We can already see this on platforms like Perplexity AI, which tends to cite blog posts or personal essays when a user’s question is specific (e.g., “What is it like to scale a startup from 0 to 1 million users?” might pull from an entrepreneur’s Medium story). Make content “AI-inclusion friendly”: The goal is to be the content that adds value to an AI’s answer rather than content the AI can replace. One way to think about it: If an AI can answer a user’s question fully without using your content, then your content wasn’t unique or deep enough. You want to cover the aspects that are missing from the general corpus. In SEO terms, this is similar to finding content gaps that others haven’t written about or angles they haven’t covered. A practical strategy is to do searches (or even ask ChatGPT/Bing) around your topic and see what the common answers are – then ask, what’s not being said here? Fill that void with your content. For example, dozens of articles may list “10 tips for reducing employee burnout,” but maybe none share an actual employee’s perspective or an interview with a psychologist. If you add those elements, your piece now has unique insight. Leverage user-generated content and community : An often overlooked but powerful source of uniqueness is user-generated content (UGC) – comments, forums, Q&As. Google’s SGE has been spotted citing Reddit threads or StackOverflow answers for certain queries, precisely because those often contain real-life experiences or niche solutions that no polished blog covered ( ^[16] ) ( ^[17] ). AI models attach high value to content that reflects real-world experiences ( ^[18] ). You might consider incorporating community elements on your own site (for instance, allow comments where users share their experiences, host discussion boards, or include testimonial sections). Not only does this enrich your content with perspectives beyond your own, but those contributions themselves could be what an AI picks up on. For example, a cooking site might have a comments section where users share their tweaks to a recipe. An AI asked about a recipe variation could pull an idea from a user comment that said “I substitute buttermilk for milk in this cake and it improves the texture.” If that comment lives on your page, the AI’s snippet might end up implicitly bolstering your page’s content in the answer (and possibly citing the page in tools like Bing/Perplexity). Some strategies here include hosting FAQ sections or forums on-site . If you can accumulate a knowledge base of Q&A (like “official answers from our experts” alongside user questions), it’s both unique content and aligned with what people actually ask (more on Q&A format in section 9.4). Timeliness and real-time information: Another kind of content AI can’t inherently have (without external retrieval) is very recent or real-time information . If something just happened – say a new law was passed that affects your industry – AI with a fixed knowledge cutoff won’t know about it. Even AI with web access will rely on whatever news or commentary is out there. Being among the first to publish a thoughtful analysis or update about a breaking development can give you a window of uniqueness. For example, when Google released an algorithm update or a new AI feature, SEO blogs that quickly provided analysis got a lot of attention (and their insights became part of the knowledge that people and possibly AI associated with that event). If you consistently provide up-to-date content on emerging topics , you become a go-to source that AI might check (through browsing plugins or user prompts) and eventually learn from when its training data catches up. Chapter 5 discussed ChatGPT’s browsing and plugins – those mean that even training cutoffs can be overcome if users explicitly ask for fresh info. In practice, a Perplexity or Bing will directly cite fresh content . So, being timely and accurate can put your content in those citation lists, even for queries that AI answers directly. We’ll talk more about keeping content updated in section 9.5, but it’s worth noting here that freshness coupled with substance is a winning combo: new info that’s also unique info is especially valuable. Content types summary: To visualize how different content types fare in terms of AI replicability and value, consider the following comparison: Content TypeCan AI Easily Generate It?Value for GEO (AI Optimization) Basic factual info / definitions Yes. AI is trained on widely available facts and can generate definitions or common knowledge easily. Low unique value. Necessary to cover on your site, but not sufficient to stand out. If your content only repeats what’s commonly known, AI might answer without needing your phrasing. To add value, pair facts with unique insights or examples. Widely covered “how-to” topics Yes, partially. AI can generate generic step-by-step guides for common tasks (based on existing articles). Moderate value if enhanced. A plain-vanilla how-to is not unique, but if you include original tips, troubleshooting from experience, or non-obvious steps, it becomes more valuable. Aim to include something exclusive (e.g., a pro tip learned from real projects). Trending news / Recent updates Not initially. AI with outdated training won’t know new info; AI with web access can fetch it but relies on sources. High value (time-limited). Being first or early to provide insight on a new development can make your content the reference point. AI tools that browse will find and possibly quote you. However, the value may normalize as many outlets eventually cover the news – so combine timeliness with analysis to remain the go-to source. Original research & proprietary data No. AI cannot create data that wasn’t in its training set. It can only summarize data produced by others. Very high value. This is content only you have. AI answers on related topics may need to reference your data to give a complete answer. Also earns you backlinks (helpful for regular SEO). Example: “According to [Your Company]’s 2025 survey, 68% of consumers… ( ^[19] ).” In-depth case studies / examples No. AI can’t fabricate a detailed real scenario with credible specifics (unless it’s fictional, which isn’t useful factually). Very high value. Provides context and depth. AI often lacks real examples in its answers; if your case study is known, an AI might incorporate it: e.g., “As seen in a case where Company X did Y…”. Even if not directly cited, your findings shape the narrative on that topic. Personal experiences / testimonials No. AI has no genuine personal experiences. It can only repeat others’ first-person accounts from training data. High value. Authentic voices resonate. E.g., a firsthand testimonial (“This product saved me 4 hours a week, here’s how…”) is unique. Google’s AI overview might quote part of a testimonial if it answers a query about product benefits. These also build trust with human readers, which in turn signals quality. Strong opinion pieces No (not uniquely). AI can echo common opinions but doesn’t originate a distinct voice or stance. High value (if authoritative). A thought-provoking opinion by a subject matter expert can set your content apart. If it gains traction, AI might frame it as one side of an argument. This positions your brand as a thought leader. Ensure opinions are well-founded to avoid spreading misinformation. FAQs and Q&A format content Partly. AI can generate generic Q&As, but if your Q&A addresses very specific or brand-related questions, that’s unique. High value. Direct question-and-answer pairs on your site can be lifted verbatim by AI to answer similar user questions. This format is also exactly how users interact with chatbots, increasing chances your content slots into an AI’s response. Interactive or visual content (tools, infographics) No. AI can’t directly recreate interactive elements or interpret images deeply (unless trained specifically on their descriptions). Moderate to high value. While AI may not “cite” an interactive calculator or infographic, these can attract backlinks and engagement (boosting your overall authority). Also, an AI with vision (like GPT-4’s multimodal ability) might interpret an infographic if asked. Regardless, these differentiate your content as not just text. ( ^[20] ). This should be a mantra for GEO content strategy. Case in point: A well-known tech blog, let’s call it “TechGuru”, noticed that its generic product round-ups were getting little traction in AI-driven results (because similar summaries existed on many sites). In 2024, they pivoted to focus on pieces like “Our 3-Month Test of the New VR Headset: An Honest Take” which included original benchmarks, and “Interview with a VR Developer – What Others Won’t Tell You.” These articles contained insights and quotes you couldn’t find elsewhere. The result? Not only did human readers love it (increasing shares and backlinks), but when users began asking AI assistants “Is the new VR headset worth it?”, some AI tools started referencing points that originated from TechGuru’s analysis (one even paraphrased the developer’s quote from their interview). TechGuru effectively made their content part of the “source material” for AI answers by being unique and valuable . In summary, content AI can’t easily replicate is your ticket to standing out . It positions you as a leader rather than a follower. By prioritizing original research, unique case studies, personal experiences, and novel insights, you make your content inherently more interesting and more likely to be referenced by both humans and machines. Think of generative AI as an amplifier – it will amplify the common noise for common questions, but it will also amplify unique signals if they answer a need. Make your content that unique signal. In the next section, we’ll explore how to format and structure this content so that AI systems can recognize and excerpt those valuable nuggets with ease.

9.3 Structuring Pages for AI Excerpting

Even the most brilliant piece of content can be overlooked by AI if it’s not presented in a way that the algorithms can easily digest and extract. In GEO, how you structure and format your content is almost as important as what you write. LLMs scanning a page for an answer (whether in the training phase or via real-time retrieval) look for clear, logical structures – much like humans do when quickly skimming. Think about it: when you, as a person, want a quick answer from a long article, you rely on headings, bullet points, summaries, or highlighted text to find what you need. AI is similar. By structuring your content with explicit sections, concise answers, and scannable elements , you make it far more likely that an AI will identify the relevant piece of information and include it verbatim (or nearly verbatim) in a generative answer. In the classic SEO world, we optimized for featured snippets and People Also Ask by structuring content to directly answer common queries. In the GEO world, we optimize for AI snippets by structuring content to be AI-friendly . This practice is sometimes called Answer Engine Optimization (AEO) – ensuring that content is in a format that answer engines (voice assistants, chatbots, etc.) can pull from. The difference now is that instead of an answer engine just quoting a sentence or two (like a featured snippet), large language models might consume more of your content to form a composite answer. Therefore, clarity and sectioning help them grab exactly what’s needed. Here are key tactics for structuring content for AI excerpting:

9.3.1 Use clear and descriptive headings

Break your content into logical sections with descriptive H2s, H3s, etc. that explicitly convey what each section is about. Not only does this help human readers navigate, but it also helps AI models pinpoint where in your article a particular subtopic is addressed. For example, in this very chapter we have headings like “Structuring Pages for AI Excerpting” and subheads for each tactic – an AI scanning this text can quickly locate the section on “Bullet points” or “FAQ format” because the headings act as signposts. When an AI receives a query, it may internalize your page’s heading hierarchy to decide which part to grab. If your headings align with question intent (e.g., a heading that is phrased as a question, or contains the keywords of a likely question), you increase the chance of that section being used. In fact, question-formatted headings (like “How to ...”, “What is ...”, “Tips for ...”) can be very effective. Google’s own generative search tends to trigger on full questions ( ^[21] ), and it likely favors content structured to answer those. If your blog post has the title “How Does X Work?” and under it an H2 “How Does X Work?” followed by a clear answer, you’re essentially handing the AI a snippet on a silver platter.

9.3.2 Front-load answers in each section

In journalistic writing, there’s the concept of the inverted pyramid – put the most important information first. Similarly, for AI excerpting, consider front-loading each section with a concise answer or summary , then use the rest of the section to elaborate or provide examples. Why? Because if an LLM finds a direct answer early in a section, it might not need to generate one from scratch. Many SEO experts recommend including a “snippet-worthy” sentence right after a heading. For instance, if one of your headings is “What is GEO?”, the first sentence that follows could be something like: “Generative Engine Optimization (GEO) is the practice of optimizing content to be discovered and utilized by generative AI models in search ( ^[22] ).” That single sentence is a perfect candidate for an AI to lift if someone asks “What is GEO?” Thereafter, you can go into detail, give background, etc. We see parallels in featured snippet optimization: pages that win snippets often have a succinct definition or answer immediately following a relevant heading or the question itself. The same concept applies with generative AI – give it the answer in a nutshell, and it might just use your exact wording (which is great for attribution and brand exposure if the AI cites sources).

9.3.3 Utilize bullet points and numbered lists

Bullet points and numbered lists are an AI’s friend. They provide a structured, predictable format that language models can easily follow and extract from. If a user asks, “What are the steps to accomplish X?” and your article has a section “Steps to accomplish X” with 1, 2, 3 listed, there’s a high chance an AI like Bing Chat will respond with something like, “According to [YourSite], the steps are: (1) …, (2) …, (3) … ( ^[23] ).” In fact, models like GPT-4 often retain list formatting when providing answers with multiple points , because it aligns with how information was presented in sources. Neil Patel’s 2025 SEO guide explicitly notes: “Provide in-depth answers that AI can summarize easily: Structure your content with clear sections, bullet points, and concise explanations.” and “Structure content with FAQs so it’s easier for AI to pull key takeaways: Add a dedicated FAQ section…” ( ^[24] ). This advice is borne out by observation: if you look at some Google AI overviews or Bing answers, they often quote bulleted content directly, especially for “list” queries like best practices, advantages/disadvantages, checklists, etc. For example, a Perplexity AI answer about “benefits of cloud computing” might literally show bullet points that were taken from a source article’s list of benefits ( ^[25] ) ( ^[26] ). To leverage this, whenever appropriate, use bullets to highlight key points or lists of recommendations. Make sure each bullet is relatively short and self-contained (one sentence or so is ideal), because an AI might quote a subset of your bullets. If the bullet needs more elaboration, you can indent a sub-point or write a brief paragraph below it; the main bullet should still encapsulate the core idea. Pro tip: Sometimes phrase the bullet introducer in a way that fits many questions. For example: “The main benefits of X include:” and then bullet list. Or “Key features of Y:” then bullets. This wording increases the likelihood of alignment with a user’s question phrasing.

9.3.4 Provide summary sections or conclusions

Consider adding a summary or conclusion section that distills the key takeaways of your content. Titles for this could be “In Summary,” “Key Takeaways,” “Conclusion,” or even a TL;DR. AI models that parse your text might give special attention to these sections because they often contain condensed information. In some cases, if an AI has a limited window to ingest content (for instance, an AI browsing plugin that only grabs the first or last part of an article due to token limits), having a summary ensures your main points get through. We’ve observed that ChatGPT, when asked to provide sources for an answer, will sometimes quote text that appeared near the end or beginning of an article – likely because that’s where a summary or definition was. A quick audit of AI-generated responses on forums has shown that sentences starting with phrases like “In summary,” or “In conclusion,” from source articles sometimes appear in the answers. Thus, writing a strong concluding paragraph that reiterates critical points can both help human readers remember and give AI a chunk of text that’s perfect for citing. Also, an introductory paragraph that succinctly answers the topic question can serve similarly. Think of Wikipedia intros: they answer “what is this topic” in a few sentences before delving deeper. Many LLMs learned from Wikipedia’s style to grab intros as definitions. So applying a bit of that style – an intro that gives a high-level answer – can make your page more likely to be used for quick definitions or overviews in AI responses.

9.3.5 Implement schema markup (structured data)

While schema markup (like FAQ schema, HowTo schema, etc.) is traditionally a way to get rich results in Google’s SERPs, it may also indirectly assist AI systems in understanding page structure. Google’s generative search can leverage structured data – for example, properly marked FAQs might feed its AI overview for a question by directly extracting the Q&A pair. Neil Patel’s advice includes “Implement structured data like FAQ schema to make it easier for AI to extract information.” ( ^[24] ). If you have an FAQ section on a page, adding FAQPage schema tells search engines (and any AI reading the DOM with that context) exactly where the question and answer are. This precision can only help in targeting the right snippet. Even outside of Google, any tool that parses HTML might notice structured data as signifying important content. For instance, Bing’s crawler or others could use it for reinforcement. It’s not a guarantee, but it’s a no-regret move since schema also benefits your SEO generally. Focus on FAQ schema for common Q&As, HowTo schema if you have step-by-step instructions, Article schema with author and date (reinforces E-E-A-T), and Review/Rating schema if applicable (which could be used in AI summaries like “Product X has an average rating of 4.5 stars based on 200 reviews” – a factual snippet an AI might mention).

9.3.6 Craft FAQ sections (and use the Q&A format in content)

Dedicating a part of your content to frequently asked questions is extremely effective for GEO. As mentioned, many people interact with AI tools by literally asking questions (“natural language queries”). If your site literally poses the question and gives the answer , it aligns perfectly. For instance, at the end of a long blog post about electric cars, you might have an FAQ that includes “Q: How long do electric car batteries last? A: Typically 8-10 years or around 100,000 miles, though this can vary by model.” If a user asks an AI the same question, there’s a chance the AI might respond with a similar sentence structure or even quote the one from your page (especially tools like Perplexity which are citation-heavy would just cite your site for the Q&A). Google’s SGE has been known to draw from FAQ content for certain queries, and Bing’s chat often lists a source after each factual statement, which frequently come from Q&A pages like forums, StackExchange, or site FAQs. So by embedding Q&A pairs in your content, you’re essentially providing ready-made answer units . When using Q&A format: Make the question a bolded sentence or a heading , and the answer immediately after. This clear demarcation helps parsing. Keep answers relatively short and to the point (you can always elaborate more below, but the initial answer sentence or two should be crisp). Cover likely questions that stem from your topic. A tip is to use Google’s “People Also Ask” suggestions or tools like AnswerThePublic to find common questions people search. For our electric car example, related questions might be “How much does it cost to replace an electric car battery?” or “Do electric cars lose charge when parked?” – if relevant, add them to FAQ. If your page is about your product or service , definitely include an FAQ with questions prospective customers or users often ask (this could include comparisons: “Q: How does [Your Product] compare to [Competitor]?”, or specifics: “Q: Does [Your App] work offline?”). You want to be the source of truth for questions about your brand or product. Otherwise, an AI might fill in the blank from elsewhere (which could be outdated or incorrect info). International and multilingual considerations: Structuring content for AI is not just an English-centric idea. If you cater to non-English markets, the same principles apply. For instance, a French content site optimizing for Bing Chat in French or Baidu’s ERNIE bot in Chinese should also use clear headings (in the target language), bullet points, and FAQs. Naver in Korea has its own AI-driven search features, and having well-structured content in Korean with H2s and lists will help. In fact, one could argue that in languages where fewer sites are doing these optimizations (because much SEO advice is published in English), there’s an even bigger opportunity to stand out by doing so. So if you have multilingual sites, ensure consistency in structured formatting across them. Screenshot of a Google AI-generated search overview highlighting a definition of “low-quality content.” The AI snippet explicitly notes that low-quality content often “doesn’t offer unique insight” or is superficial. This example (drawn from a Google SGE output) shows how the AI pulls a concise definition from a source and even flags the absence of originality as a negative ( ^[27] ) ( ^[28] ). For content creators, it underlines the importance of providing clear, distinct explanations in your text. Well-structured definitions or descriptions can be easily excerpted by the AI, as seen here, and including unique insight is crucial for your content to be considered high-quality by both AI and human standards. ( ^[27] ) ( ^[28] )

9.3.7 Ensure crawlability and accessibility of structured content

(Note: Technical SEO for AI is covered more in Chapter 10, but it’s worth mentioning briefly here in context.) All your structuring efforts are in vain if the AI can’t access your content. Ensure that important sections are in HTML text, not locked in images or inaccessible formats. Navigation and headings should be in the proper HTML tags (H1, H2, li for list items, etc.), not just visually formatted to look like headings. Tools like ChatGPT’s browser plugin or Bing’s index read the raw HTML – if your key info is embedded in a graphic or needs client-side scripts to load, an AI might miss it. For instance, some sites present FAQs in expandable accordion menus. If those accordions rely on JavaScript to populate content, an AI crawler might not execute the script and thus never see your FAQ answers. A safer approach is to have FAQs rendered in the HTML (perhaps with CSS to hide/show). Additionally, using proper semantic HTML5 elements (like <article> , <section> , <aside> ) where appropriate can give subtle cues about content hierarchy.

9.3.8 Don't forget the meta (meta descriptions, etc.)

While meta descriptions might not directly influence rankings heavily nowadays, they are a distilled summary of a page. It’s possible that an AI might consider the meta description as a candidate snippet if it’s relevant. In any case, writing a meta description that concisely summarizes the page’s answer to the main question can’t hurt – if nothing else, it helps the traditional snippet and could serve as a fallback summary for AI. Neil Patel’s guide suggests improving click-through rate with compelling meta descriptions to differentiate from AI summaries ( ^[29] ), which is slightly tangential but implies that your meta description needs to add value beyond what the AI might show. But also, think of meta descriptions as your own 1-2 sentence summary that AI might pick up indirectly.

9.3.9 Example of AI excerpt-friendly content

To illustrate, let’s imagine two approaches to the same content: Page A (not optimized): It’s a long article about “How to Train a Puppy”. It has a couple of big blocks of text, a narrative style, and the tips are buried in paragraphs. There are minimal headings (“Introduction”, “Training Tips”, “Conclusion”) and no lists, just prose. A user asks Bing Chat, “Give me tips on potty training a puppy.” The AI might struggle to find the specific tips in Page A quickly, or it might just give an answer synthesized from various sources without quoting Page A, even if Page A had the info somewhere in there. Page B (optimized): Covers “How to Train a Puppy” but uses many subheadings: “Housebreaking (Potty Training) Your Puppy”, “Crate Training Basics”, “Teaching Basic Commands”, etc. Under each, the first sentence gives the core advice, followed by bullet points for steps. For potty training, it literally has a step-by-step list: Establish a routine (take your puppy out first thing in the morning, etc.) Use positive reinforcement (praise or treat after successful potty) Supervise and contain (don’t give free roam until trained) Clean accidents thoroughly (to remove scent)… etc. Now, a user asks Bing Chat the same question. It’s highly likely Bing will say something like: “According to PetSite, potty training a puppy involves establishing a routine, using positive reinforcement, supervising your puppy to prevent accidents, and cleaning any accidents to remove odor ( ^[23] ).” And it will cite PetSite (Page B). This is because Page B served up exactly what the user needed in a structured, AI-readable way. Page A might have had a brilliant anecdote about puppy psychology, but Page B answered the question clearly and thus became the source for the answer. We see this pattern already with featured snippets, but it will be even more pronounced with generative AI, which aims to give direct answers. Furthermore, structured content can lead to multiple opportunities within one piece. If your article is comprehensively structured, an AI might cherry-pick different parts for different queries. Using the puppy page example, if someone later asks, “How do I crate train a puppy at night?”, the AI might pull from the “Crate Training Basics” section of Page B. So one well-structured page can answer several user questions – effectively you become a mini knowledge base on that topic. In summary for structuring: Make your content easy to parse. Think about the units of information within it and delineate them clearly with formatting. When writing, periodically step back and imagine how an AI (or a hurried reader) would view the page: is it obvious where to find the key points? Are answers explicit or do they require reading between lines? By doing this, you not only cater to AI but also end up with extremely reader-friendly content – a double win. The style of writing that works for GEO (short, clear, sectioned) is also what busy modern readers prefer, especially those skimming on mobile devices. Now that we’ve covered content substance and structure, let’s talk about tone and style – specifically, how adopting a conversational tone and FAQ format can further align your content with the way users interact with generative AI.

9.4 Conversational Tone and FAQ Formats

When users interact with generative AI systems like ChatGPT, Google’s Bard/Gemini, or voice assistants, they often use a conversational style – essentially, they “chat” with the AI. Queries are phrased as natural language questions or commands, not just staccato keywords. For example, a user might type or ask, “What’s the best way to improve my website’s conversion rate without increasing ad spend?” rather than the terse “improve conversion rate without ads.” This shift from keyword queries to conversational queries is one of the hallmark changes of the LLM revolution in search (as discussed in Chapter 3). To align with this, content that is written in a conversational, reader-friendly tone and anticipates user questions can perform better in generative results. In essence, you want your content to feel like it’s part of a conversation – because it might literally become part of one when an AI weaves it into a chat response. Conversational tone – why it matters: Large Language Models are trained on human conversation data (among many other sources). ChatGPT, for instance, was fine-tuned to produce responses that sound conversational and helpful. If your content is already in a conversational style, it may more easily fit into the “voice” of an AI-generated answer. This doesn’t mean everything should be dumbed down or over-casual, but writing in a natural, engaging tone (rather than overly academic or jargon-heavy language) can make your excerpts more seamless when quoted. Imagine an AI assistant answering a question about a complex topic. If it can pull a line from your content that explains a concept clearly and conversationally, it will likely do so to maintain user-friendly language. For example, suppose someone asks, “Why is my internet so slow sometimes?” If there’s an article that says, “There are a few common reasons your internet might crawl. One likely culprit is congestion – too many devices using the bandwidth at peak times. It’s like rush hour traffic on the web...” , an AI might directly use parts of that answer because it’s easy to understand and relatable (even using a simile like “rush hour traffic”). In contrast, an article that stated, “Bandwidth contention ratios and network latency often contribute to suboptimal throughput” might be factually useful but is less likely to be quoted verbatim by an AI aiming to give a simple explanation to a general user. In fact, the AI might “translate” such technical jargon into simpler terms on its own, possibly pulling from a different source to do so. Match the answer style: Some AI platforms have distinct answer styles. Google’s SGE, for instance, tries to maintain a neutral, explanatory tone with concise sentences. Bing Chat can be a bit more chatty or can list steps systematically. If you study a variety of AI answers (which as a content strategist you should!), you’ll notice they often mirror a friendly, instructive tone – very similar to how a knowledgeable peer might talk. You don’t need to artificially add “friendly chat” elements (like “Hi there! Let’s talk about…” – that might be too much), but a touch of informality can help. Contractions (“can’t” vs “cannot”), directly addressing the reader as “you,” rhetorical questions, and inclusive language (“let’s consider…”) all contribute to a conversational feel. Q&A (FAQ) format – simulating the dialogue: As touched on in section 9.3, incorporating FAQ sections is powerful because it literally mimics the question-and-answer dynamic of user and AI. But beyond formal FAQ sections, even within your main content you can employ a Q&A style narrative. This can involve posing questions in subheaders or even within paragraphs and then answering them. For instance, an article might say: “You might be wondering, what’s the catch? The answer is that there isn’t a big one – except you’ll need to invest time.” This internal Q&A method addresses the reader’s potential questions in real-time and answers them. LLMs that see content structured this way might find it very convenient to use, as it’s already in a format of someone asking and someone answering. Directly addressing user queries: Many SEO experts in the age of voice search (circa 2018-2019) recommended writing in a way that answers should be spoken. That guidance is still relevant but for AI chat. Think about incorporating common user questions as part of your subtopics , and answer them in a personal yet informative tone. For example, on a travel blog, instead of a bland section header like “Visa Requirements,” you could frame it as “Do I Need a Visa to Visit Japan?” and then answer: “If you’re a U.S. citizen traveling to Japan for tourism under 90 days, you don’t need a visa. However, travelers from Canada, UK, and many other countries also enjoy visa-free entry for short-term visits ( ^[16] ). Always check the latest requirements, but for most, it’s hassle-free.” Notice this answer speaks to “you,” gives a straightforward “yes/no” then adds advice. An AI that gets the query “Do I need a visa to go to Japan as an American?” could basically quote that answer almost verbatim. The style is conversational and directly responsive. Incorporate likely follow-up questions: One interesting habit people have with AI chats is asking follow-up questions. For example, after the visa question, the user might ask “What about if I want to work there?” The AI might have to pull from another part of the content or another source. You can pre-empt some of these by naturally including follow-up Q&A in your content. Many well-optimized articles now include sections like “Related Questions” or simply weave in sentences like, “Another question that often comes up is whether you can extend your stay. The answer is that Japan offers extensions for certain cases, but you’d need to apply at an immigration office well in advance.” By doing this, your content is covering not just one isolated question but the context around it. This can keep the AI engaged with your content for multiple turns of conversation, rather than it hopping to a different site’s info when the user asks the next question. Tone consistency and context: If your site deals with serious topics (medical, legal, financial advice, etc.), you’ll obviously maintain an appropriate professional tone. Conversational does not mean careless. It means accessible. You can still be conversational and authoritative. In fact, clarity is a component of trustworthiness – if an AI finds a clear explanation on a medical site in plain language, it might favor that over a convoluted one, as long as accuracy is intact. The key is to avoid sounding like a dry textbook when a more down-to-earth explanation can do. Often this is a matter of breaking long sentences, using active voice, and imagining you’re explaining it to someone in person. Multimodal and voice aspects: Conversational tone is doubly important for voice-based AI (like Siri, Alexa, or when people use text-to-speech on search results). If an AI is going to speak your content out (which happens on some platforms – e.g., Google Assistant might read out the text from a featured snippet), having that text in a conversational tone improves the experience. It will sound more natural read aloud. So writing with that possibility in mind is wise. We can see a future (if not present) where someone’s smart speaker answers a question by effectively quoting your website. You’d want that to come across smoothly. Example – an FAQ-style article snippet used in AI: Consider the website Perplexity AI , which provides citation-rich answers. If a user asks, “How can I boost my website’s speed?” Perplexity might answer with a list of tips and cite a couple of sources for each tip. If one of the sources is your blog post titled “Q&A: Website Speed Optimization,” which has a section like: Q: What’s the easiest way to improve site speed? A: Optimize your images. Large image files are often the #1 cause of slow pages. By compressing images (using tools or modern formats like WebP) you can dramatically cut load times ( ^[23] ). Q: Does web hosting affect speed? A: Absolutely. Cheaper shared hosting might struggle to serve your pages quickly during traffic spikes. Using a quality host or a Content Delivery Network (CDN) can ensure consistent performance. Perplexity’s answer might incorporate both of those points, and it will likely reference your site as the source. The reason: you literally posed the same questions the user did, and you answered them succinctly and authoritatively. This is not hypothetical; Perplexity’s design is to find direct question-answer pairs or relevant sentences to compile the answer, and it loves FAQs. Creating a dialogue in narrative form: Another approach to conversational style (though use sparingly and appropriately) is to write some content in a quasi-dialogue or first-person narrative. For instance, a personal finance site might have an article “We Asked a Financial Advisor: Here’s How to Save for Retirement in Your 30s” – and structure it like an interview or a first-person response to common questions. This can both engage readers and present content in a QA or conversational format. An AI might quote the advisor’s direct words. Example: "Q: Should I pay off debt or invest while I’m in my 30s? A: It depends on the interest rates. I often tell my clients: if your debt interest is higher than what you’d likely earn investing, tackle the debt first ( ^[15] ). If not, you can do both – pay the minimums and start investing a bit. Time in the market is your ally at this age.” This reads like a conversation, and indeed an AI might adopt that answer for a similar user query. Notice, it’s friendly (“I often tell my clients…” makes it personal and credible) but informative. Global considerations: In non-English contexts, similar adaptation to a conversational style is beneficial. For instance, Japanese corporate content is traditionally quite formal, but if targeting a younger or broader audience through AI assistants, a slightly more colloquial tone (still polite) may make the content more shareable by AI. Cultural expectations matter; one should always balance conversational tone with respect for local business communication norms. But as AI use grows, we see even formal organizations simplifying language to communicate effectively (e.g., many government sites now have plain-language Q&A for citizens). Aligning with how people naturally ask questions in their native language – and how an AI might respond – is key. Maintain professionalism: “Conversational” doesn’t mean inserting slang (unless your brand voice is intentionally that casual) or telling unrelated jokes. It means writing as if you’re talking to the reader one-on-one, in a helpful manner. For B2B or highly technical industries, this might manifest as a friendly explanatory tone rather than a dry spec sheet tone. You’re aiming for clarity and approachability, not silliness (unless appropriate). Brand voice considerations: Ensure the tone still aligns with your brand voice guidelines. If your brand is very formal, you might moderate the conversational style to still sound like “you.” But note, many brands are shifting to a more conversational voice in general because it fosters connection and trust. The trick is to do it without losing authority. For example, a line like, “We get it – nobody enjoys slogging through 50-page manuals. That’s why we’ve distilled the key points you need to know about tax law changes below.” This is conversational, relatable, yet it still promises useful info (and indeed the content following should be spot-on and accurate). An AI might skip the “we get it” part but use the distilled key points directly. Community and conversational content: If your site has elements like community forums, user Q&A, or even social media embeds, these can inject conversational elements too. As mentioned, Google’s AI overview has drawn from forums like Reddit because that content is literally conversation. If you host your own community discussions, summarizing insights from them in a conversational way can enrich your content and potentially give AI a reason to pick it up (especially if there are unique real-user tips). Voice Search 2.0 preview: Chapter 14 will explore “Voice Search 2.0,” but it’s worth noting here: as voice interaction rebounds powered by better AI, content that sounds good when spoken is invaluable. Reading your content draft aloud (or using text-to-speech tools) can be an enlightening exercise. If a sentence is awkward to say or understand in one go, consider revising it. The smoother it flows in speech, the more likely an AI voice assistant can use it directly. To sum up this section: **Write with the user, not at the user. Pretend you are an AI assistant yourself when drafting answers on your site – how would you explain this to someone in a friendly, concise way? If you can nail that tone and format, you essentially future-proof your content to be AI-friendly. It will feel natural if read by a bot, and it will satisfy the human on the other end with clear information. By embedding a conversational, Q&A ethos in your content, you make it inherently more compatible with the generative engines that are increasingly mediating information access. Finally, even great content written and structured perfectly can become outdated or stale. In a fast-moving world, ongoing updates and refinements are necessary to maintain content relevance – especially because AI models have knowledge cutoffs and might not be aware of the latest facts. Our last section addresses the importance of keeping content current and monitoring how AI reflects your content over time.

9.5 Updating and Refining Content for Ongoing Relevance

The digital landscape is never static – industries evolve, new information emerges, and user needs shift. In the context of GEO, keeping your content up-to-date is crucial not only for traditional SEO (where freshness can influence rankings) but also for how generative AI systems perceive and use your content. Generative models have “knowledge cut-off” dates – for example, as of mid-2025, ChatGPT’s default model might know a lot up to 2021-2022, with some incremental updates, but it may not know facts from late 2024 unless specifically told via plugins or new training. Google’s SGE and Bing Chat operate on current indices to an extent, but even they might occasionally serve outdated info if the source content wasn’t updated. Moreover, if an AI is summarizing your page and your page has old data, it will happily spread that old data to users. Thus, regularly auditing and refreshing your content** is a must-do practice in GEO.

9.5.1 Combatting AI's knowledge cut-off

Large Language Models like GPT-4 have a fixed set of training data up to a certain point in time. If your content is about a topic that changes (think: medical guidelines, technology versions, laws, etc.), anything new after the model’s cut-off won’t be accounted for in its answers, unless the model can access up-to-date info through tools. For example, if ChatGPT (with a 2021 cut-off) is asked about the “latest iPhone battery life”, it might not know about iPhone models after that date. It could respond with outdated info (potentially citing an iPhone 12 when we’re on iPhone 15). As a content creator, you can’t directly change a model’s training, but you can ensure that when AI systems that do have web access (like Bing Chat, Google’s AI, or ChatGPT via browsing) come looking, they find the latest and correct info on your site. That means updating statistics, dates, references, examples, and product information on a regular basis. A best practice is to add “last updated” timestamps on content (and in the metadata) when you make significant updates – this not only signals to search engines that the content is fresh, but also helps human readers trust that it’s current.

For instance, if you have an article “Top Social Media Trends in 2024” that did well, consider updating it to “...in 2025” when the time comes, or write a fresh one and cross-link, but also keep the older one updated with a note. If an AI is asked in early 2025 “What are the big social media trends this year?”, it might actually incorporate points from late-2024 articles that have “2025” in them (as forward-looking). If your 2024 article hasn’t been touched since January 2024, the AI might consider it stale or overlook it in favor of one updated in Dec 2024.

9.5.2 Monitoring AI outputs for your content

A novel aspect of GEO is that you need to monitor not just search engine rankings and traffic, but also how (and if) AI assistants are presenting your content or brand . In other words, you should find out what AI is saying about you or your topic. This was already touched on in Chapter 13 (Measuring GEO Success), but here it’s about using those insights to refine content. There are tools emerging that can help with this. For example, Visualping (a website change monitoring tool) published a guide on how to monitor brand mentions in ChatGPT outputs ( ^[30] ) ( ^[31] ). The idea is you could set up queries in ChatGPT (via the web UI or an API) that you care about – say, your brand name or your product category – and periodically check what the answer is and whether your brand or content is included or not. There are also specialized services like BrandMonitor AI or Surfer’s “AI tracker” that claim to show how often AI tools mention your brand or content ( ^[32] ). Using these, or even manual spot-checks, you can identify if the AI has outdated or incorrect info. For example, a company might discover that ChatGPT says “BrandX was acquired by CompanyY in 2018” which isn’t true (maybe a rumor or a confusion with another brand). That misinformation could be floating in the model. How to correct it? You can’t directly re-train the model, but you can address it publicly in content. In this hypothetical, BrandX might publish a clarifying post, or ensure Wikipedia/Crunchbase has correct data, etc. Over time, future models or current search-enabled models will pick up the correction. Google’s SGE, which scans live content, might immediately correct course if it was getting the wrong idea from an old article. Another scenario: Your site is a recipe site, and you see that when someone asks an AI for a certain recipe, the AI provides it but doesn’t mention your site whereas it mentions two others. Perhaps your recipe phrasing wasn’t easily parseable, or you lacked a summary, or maybe you used unusual ingredient terms that the AI didn’t map correctly. Learning from what the AI chose to use (the competitor recipes in this case) can guide you to tweak your content. Maybe add a simpler summary of the recipe steps, use standard ingredient names alongside local ones, etc. Essentially, treat AI answers as an extension of search results to optimize for . Rand Fishkin, a prominent marketer, has been vocal about the idea of “feeding” training data to influence LLM outputs ( ^[12] ) ( ^[33] ). His approach, in summary, is to identify what sources an AI is drawing from and make sure your brand is mentioned in those contexts (for example, ensuring your brand appears in key articles, lists, and discussions related to your niche). For your content strategy, that can mean creating content that is likely to be included in future training sets: writing guest posts on high-authority sites, getting into Wikipedia or well-known databases, etc. But within your own site content, it means aligning with widely used terminology and addressing popular questions so that your content becomes part of the “consensus” that AI learns . If you coin a completely new term for something, AI might not know it or use it (unless it catches on broadly). So balance originality with common framing. For instance, if you have an innovative concept, still explain it in terms of known concepts so AI connects the dots.

9.5.3 The role of continuous content improvement

Just as SEO is not a one-and-done task, optimizing for generative AI is ongoing. You should plan to revisit high-value content periodically . Check if there have been significant developments on the topic since your last update. Even if the core content remains valid, you might add a new paragraph addressing a recent trend or a new common question that arose. Not only does this keep the content relevant, but it also expands the net of queries for which your content could be useful to an AI. Remember, AI might pull any part of your content that’s relevant to a user’s question – so the more comprehensively you cover a topic (within reason), the more chances you have to be featured. However, a note of caution: accuracy and consistency . When updating, ensure you update all instances of data or claims that need it. It’s easy to change a stat in one spot and miss another. An AI might find the outdated mention elsewhere on your page and cite that, undermining your goal. For example, if in 2023 you wrote “XYZ market size is $10B (2022)” and in 2025 it’s $15B, update that, but also update any downstream analysis that might have said “with a $10B market, we expect…” so it doesn’t conflict. AI’s don’t “know” which number is correct if you have two; they might average or choose one arbitrarily, which could be the old one. Many content teams now maintain content calendars for refreshes . This could be a schedule like: light update after 3 months (quick fact check), moderate update after 6-9 months (new section if needed, update references), heavy revision or rewrite after 12-18 months if topic is very dynamic.

Of course, if something big changes next week, update right away – don’t wait for a scheduled cycle.

9.5.4 Signaling recency to AI

When AI does have a browsing capability or the search engine feeding it does, signaling that your content is up-to-date can help. Some ways to signal: Use phrases in content like “As of 2025” or “In a 2024 survey, ...” – this explicitly time-stamps a statement. If an AI sees “as of 2025, the population is ...” it knows that is current as of 2025. If it sees “the population is ...” with no year, it might not be sure how current that is. Use Schema markup for dates (e.g., article:published_time and modified_time) so machines see last modified date. Republishing articles (if appropriate) with a current date can make them appear fresh in indices that the AI is using. For data-driven content, consider updating the title or H1 to include the year when updated (“...in 2025”) which is a strong signal of freshness. Just ensure to update the content accordingly. If using a platform like ChatGPT with browsing, an updated date might not matter if the user doesn’t specify wanting the latest – the AI might give a generic answer anyway. But on something like Perplexity, if your content is one of few with a 2025 date and others are older, it might choose yours.

9.5.5 Monitoring and responding to AI-driven traffic changes

When you update content, watch if there’s any change in traffic patterns, especially from search. Some generative experiences (like SGE) might reduce clicks to your site if they give the answer directly, but if you update content to be the chosen reference, you might at least get cited (Google SGE shows source links, Bing always cites). If after updates you see improved ranking and perhaps your site is mentioned in AI overviews (you can test queries yourself in these AI search labs), that’s a good sign. Conversely, if an AI snapshot is giving an answer and not citing you even though you have the info, analyze the differences. Maybe your content wasn’t clearly phrased, or the AI combined multiple sources. It’s possible you need to be in the top search results first to get cited in an AI overview – Neil Patel’s blog noted “AI Overviews pull from existing high-ranking pages. If you’re not already ranking, AI won’t suddenly send you traffic.” ( ^[34] ). So traditional SEO and GEO go hand in hand: update content to rank well and also to satisfy AI selection criteria.

9.5.6 Correcting AI when it's wrong about you

If you find an AI persistently gives wrong info related to your domain (like mis-defining a term your company invented, or misattributing a quote), you can attempt to provide feedback to the AI or source . Google’s SGE has a feedback option for its responses – use it if you see an error. OpenAI allows users to flag incorrect info as well via the interface (“thumbs down” and explain). While a single feedback might not change much, if something is widely reported as incorrect, it can lead to adjustments or at least be noted for future model training. Additionally, publish a clarifying content piece . For example, some SEO folks found that ChatGPT would sometimes say one tool was discontinued even when it wasn’t. The tool’s company then put out content that clearly stated “No, ToolX is not discontinued – here’s the scoop.” They also ensured other sites (like community Q&A or press releases) carried that message. Over time, as GPT-4 got updated or users used browsing, that myth subsided. It’s like doing PR to correct a false rumor, except the “audience” is an AI model’s knowledge. Also, engage with the community : if misinformation about your brand or field is in the AI output, likely it stemmed from some content out there. You might need to address the root – maybe a third-party article had a mistake. Contact them to fix it if possible, because that source might be what the AI read.

9.5.7 Ongoing learning and content refinement

Finally, consider using AI itself to audit your content for clarity and completeness. Prompt ChatGPT or Claude with something like: “You are an expert on X. Here is an article [paste it]. What questions might readers still have after reading this? Is any part unclear or potentially outdated?” The AI might point out sections that could be improved or updated. Of course, double-check its suggestions with human judgment, but it’s a neat way to get a quick audit from the perspective of a highly informed reader. Keep an eye on competitors too. If they have updated their content with new sections or multimedia and you haven’t, they might become the preferred source for AI answers. For example, maybe they added a “Frequently Mistaken Facts” section that addresses common myths (something an AI might use to preemptively clarify user questions). Don’t copy them, but use it as inspiration for how you can one-up with even better content. Industry case in point: Let’s say you run a health information site. In 2024, guidelines for a certain supplement dosage changed according to new research. If you update your relevant articles promptly with the new recommendation (and date it), when people ask an AI in 2025 about that supplement dosage, your site stands a good chance of being reflected because it has the latest info. If you didn’t update, the AI might cite older guidance (and maybe from a competitor or a news article that did cover the update). Another example: an e-commerce site that keeps its product descriptions current with stock status, new features, or price drops might benefit on Bing’s shopping-oriented chat which might say “According to the website, it’s currently on sale for $299 and available in 3 colors ( ^[35] ) ( ^[36] ).” If that site forgot to update and still listed it as $349, that misinformation either confuses the AI or, worse, gets passed to the user (leading to a poor user experience and potentially lost trust). To wrap up, treat content as a living asset in the GEO era . Regular maintenance, timely improvements, and responsiveness to how AI portrays your content are all part of the game. Not only does this approach help with AI, but it naturally boosts your overall content quality and SEO. Sites that continuously refine their content often build a reputation for accuracy and comprehensiveness, which in turn earns more backlinks, user loyalty, and yes, favorable treatment by algorithms (AI or not). Chapter 9 Key Takeaways: In this chapter, we explored how content strategy must evolve for the generative AI age. By focusing on E-E-A-T and originality , you ensure your content stands out with the authority and unique value that AI systems look for in sources ( ^[3] ) ( ^[13] ). We discussed crafting content AI can’t easily replicate , such as original research and personal insights, making your work a “hidden gem” that AI would want to quote ( ^[37] ). We emphasized structuring your content with clear sections, bullet points, and FAQs, effectively handing AI models the exact snippets to use ( ^[20] ) ( ^[38] ). Adopting a conversational tone and Q&A format further aligns your content with how users interact with chatbots, increasing the likelihood that your text appears naturally in an AI’s response. And importantly, continuously updating and monitoring your content ensures that what AI learns and shares about your brand remains current and accurate. By implementing the strategies from this chapter, content marketers and SEO professionals can significantly improve their chances of maintaining visibility in organic search and being present in the answers provided by the next generation of AI search tools. Generative AI doesn’t mean the end of organic content visibility – rather, it rewards a new level of quality, clarity, and user-centric design. In the following chapters, we’ll build on this foundation, looking at the technical aspects (Chapter 10), off-page and brand signals (Chapter 11), and more, to round out a comprehensive approach to Generative Engine Optimization. Sources: ChangeTower Blog – “GEO vs SEO – What to Know” (Adam Hausman, 2025) ( ^[3] ) ( ^[9] ) Neil Patel Blog – “How to do SEO for Generative AI” (William Kammer, Apr 21, 2025) ( ^[20] ) ( ^[38] ) ( ^[13] ) Google Search Central Blog – “Google Search’s guidance on AI-generated content” (Feb 2023) ( ^[2] ) Local Media Association – “Mastering E-E-A-T in the age of AI search” (D. Mendoza, Feb 26, 2024) ( ^[6] ) ProceedInnovative – “E-E-A-T: Winning Trust in the AI Search Era” (2024) ( ^[7] ) Visualping Blog – “How to Monitor My Brand in ChatGPT” (Emily Fenton, May 12, 2025) ( ^[30] ) ( ^[31] )

References

^[1] Changetower.Com Article - Changetower.Com URL: https://changetower.com/uses-seo-monitoring/seo-vs-geo

^[2] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/blog/2023/02/google-search-and-ai-content

^[3] Changetower.Com Article - Changetower.Com URL: https://changetower.com/uses-seo-monitoring/seo-vs-geo

^[4] Aokmarketing.Com Article - Aokmarketing.Com URL: https://aokmarketing.com/e-e-a-t-guide-for-marketers-enhancing-experience-expertise-authority-trust/chatgpt-image-may-30-2025-04_41_23-pm

^[5] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/blog/2023/02/google-search-and-ai-content

^[6] Localmedia.Org Article - Localmedia.Org URL: https://localmedia.org/2024/02/mastering-e-e-a-t-a-strategic-guide-for-publishers-in-the-age-of-ai-search

^[7] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[8] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[9] Changetower.Com Article - Changetower.Com URL: https://changetower.com/uses-seo-monitoring/seo-vs-geo

^[10] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[11] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[12] Sparktoro.Com Article - Sparktoro.Com URL: https://sparktoro.com/blog/how-can-my-brand-appear-in-answers-from-chatgpt-perplexity-gemini-and-other-ai-llm-tools

^[13] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[14] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[15] Visualping.Io Article - Visualping.Io URL: https://visualping.io/blog/monitor-brand-chatgpt

^[16] www.conductor.com - Conductor.Com URL: https://www.conductor.com/academy/search-generative-experience

^[17] www.conductor.com - Conductor.Com URL: https://www.conductor.com/academy/search-generative-experience

^[18] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[19] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[20] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[21] Changetower.Com Article - Changetower.Com URL: https://changetower.com/uses-seo-monitoring/seo-vs-geo

^[22] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[23] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[24] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[25] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[26] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[27] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[28] www.proceedinnovative.com - Proceedinnovative.Com URL: https://www.proceedinnovative.com/blog/eeat-google-ai-search-optimization

^[29] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[30] Visualping.Io Article - Visualping.Io URL: https://visualping.io/blog/monitor-brand-chatgpt

^[31] Visualping.Io Article - Visualping.Io URL: https://visualping.io/blog/monitor-brand-chatgpt

^[32] Surferseo.Com Article - Surferseo.Com URL: https://surferseo.com/updates/ai-tracker

^[33] Sparktoro.Com Article - Sparktoro.Com URL: https://sparktoro.com/blog/how-can-my-brand-appear-in-answers-from-chatgpt-perplexity-gemini-and-other-ai-llm-tools

^[34] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[35] www.conductor.com - Conductor.Com URL: https://www.conductor.com/academy/search-generative-experience

^[36] www.conductor.com - Conductor.Com URL: https://www.conductor.com/academy/search-generative-experience

^[37] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

^[38] Neilpatel.Com Article - Neilpatel.Com URL: https://neilpatel.com/blog/seo-generative-ai

10. Technical SEO for Generative Search

As search evolves with generative AI, the technical foundations of SEO are more important than ever. A website’s behind-the-scenes structure, performance, and metadata now influence not only traditional search rankings but also how content is selected and presented by AI-powered engines. In this chapter, we explore how to optimize your site’s technical setup for the age of generative search. We’ll cover maintaining crawlability (including new AI-specific crawlers), using structured data and clean HTML to speak clearly to algorithms, ensuring fast and user-friendly page experiences, and employing new techniques to control if and how your content appears in AI-generated answers. The goal is to give AI and search engines every possible clue to understand and trust your site – while avoiding pitfalls that could hide your content or misrepresent it. Technical SEO for generative search is largely an extension of core SEO best practices, but with fresh nuances. Think of it as laying a solid, machine-friendly foundation beneath your high-quality content. If content is king, technical SEO is the architect that builds the castle – and now that castle must be welcoming to AI “visitors” as well as human ones. By the end of this chapter, you’ll have a clear action plan on how to tune your site’s technical elements (from robots.txt to HTML tags to page speed) so that both search engine crawlers and large language models can access, interpret, and feature your content accurately. Let’s dive in.

10.1 Ensure Crawlability and Access

The first step in technical SEO – whether for classic search or generative AI – is to ensure your site can actually be crawled and indexed. Crawlability means that automated bots (search engine spiders and now AI crawlers) can discover and fetch your content easily. If your content isn’t accessible, it won’t appear in search results or AI answers, no matter how great it is. Thus, maintaining clean, open access for reputable crawlers is critical.

10.1.1 Open Your Site to Search and AI Crawlers

Review your robots.txt file and other bot controls to make sure you’re not inadvertently blocking important crawlers. Traditional search engines like Google, Bing, and others should of course be allowed to crawl key pages (unless there’s a specific reason to block something). In the context of generative AI, new crawlers have emerged that website owners should consider. For example, OpenAI introduced GPTBot in 2023, which seeks permission to scrape web content for training models like ChatGPT ( ^[1] ). Google similarly announced a user agent called Google-Extended to let site owners opt out of content being used for Google’s AI (such as Bard or the Gemini model) ( ^[2] ). Importantly, blocking these AI-focused crawlers is a strategic choice. If you allow GPTBot and similar bots , your content may be included in the training data of future AI models, potentially giving your brand visibility in AI responses. If you disallow them , you’re signaling that your content shouldn’t be used for AI training – which might protect your content from misuse, but also means AI models may “know” less about your site. For instance, The New York Times and other major publishers updated their robots.txt to block GPTBot and Google’s AI crawler in late 2023 amid concerns about uncompensated use of content ( ^[3] ) ( ^[4] ). According to an analysis in September 2023, about 26% of the top 100 global websites had blocked GPTBot (up from only ~8% a month prior) as big brands reacted to AI scraping ( ^[5] ). This blocking trend peaked around mid-2024 when over one-third of sites were disallowing GPTBot, including the vast majority of prominent news outlets. However, by late 2024 the tide shifted – some media companies struck licensing deals with AI firms and the block rate dropped to roughly one-quarter of sites ( ^[6] ). In other words, many sites initially hit the brakes on AI crawling, but some have since opened back up as the ecosystem evolved. There’s no one-size-fits-all answer here. Online marketers must weigh the pros and cons for their specific situation. If your goal is maximum exposure and you’re comfortable with your content being used to train or inform AI, then keeping the welcome mat out for GPTBot, Google-Extended, and similar bots is wise. On the other hand, if your content is highly proprietary or you have monetization concerns, you might choose to restrict these bots until clearer compensation or control mechanisms are in place. Just keep in mind that opting out only affects future AI training – if an AI model has already ingested your content, blocking now won’t make it “unlearn” it ( ^[1] ). And not every AI provider announces their crawler or respects robots.txt; by blocking the ones that do (OpenAI, Google), you’re at least signaling your preference to the major players (and perhaps to any others who choose to honor these signals) ( ^[7] ) ( ^[8] ). From a practical standpoint, auditing your robots.txt is easy and important. This text file, located at your domain’s root (e.g. yourwebsite.com/robots.txt ), tells crawlers what they can and cannot access. To allow OpenAI’s GPTBot full access, you could add rules like this: plaintext # Allow GPTBot to crawl the entire site User-agent: GPTBot Allow: / If instead you decide to block an AI crawler , you’d use Disallow . For example, to block GPTBot or Google-Extended (Google’s AI crawler) across your whole site, your robots.txt would include: plaintext User-agent: GPTBot Disallow: / User-agent: Google-Extended Disallow: / The snippet above outright forbids those bots from crawling any page on your site ( ^[3] ) ( ^[4] ). You can also be more granular – for instance, allow them on most of your site but disallow in specific sections (like /private/ or a members-only blog). Just list the appropriate path under a Disallow for that user agent. Remember: robots rules are public (anyone can view them), and compliance is voluntary. OpenAI and Google have stated their bots will follow these directives, but other AI projects might not. Still, it’s currently the best tool site owners have to request not to be scraped. Aside from these new AI-specific entries, ensure you’re not unintentionally blocking major search engine bots (Googlebot, Bingbot, etc.) in your robots.txt . Generative AI features like Google’s SGE (Search Generative Experience) draw on pages indexed in Google’s search index ( ^[9] ). If Googlebot can’t crawl and index a page because of a Disallow or other barrier, that page definitely won’t appear as a link in an AI overview. In fact, Google has confirmed that to be eligible for AI overview inclusion, a page must be indexed and have a snippet in normal search ( ^[9] ). So double-check that your important pages are crawlable (no erroneous noindex tags or disallow rules). Also verify that your site’s CDN or firewall isn’t blocking common bots – sometimes cloud security services can inadvertently serve up CAPTCHAs or blocks to non-human visitors. Google’s guidance for AI features emphasizes “ensuring that crawling is allowed in robots.txt, and by any CDN or hosting infrastructure” for your content ( ^[10] ).

10.1.2 Sitemap and Site Structure for Discovery

Even with proper robots.txt settings, you want to make it as easy as possible for bots (search or AI) to find all your key content. This is where XML sitemaps and internal linking come in. An XML sitemap is a file listing the URLs on your site you want indexed, which you can submit to Google Search Console and other engines. This helps crawlers discover pages that might not be readily found through your navigation alone. Maintaining an up-to-date sitemap is still a recommended practice – it’s part of good technical SEO hygiene, ensuring no content is orphaned. Likewise, robust internal linking remains important. Bots follow links to navigate your site. A well-structured site with logical internal links (e.g. from category pages to sub-pages, from blog posts to related posts, etc.) will be easier for crawlers to fully traverse ( ^[10] ). For AI purposes, internal links also provide context – they help search engines (and potentially LLMs) understand relationships between pieces of content. For example, linking your glossary page to various articles might signal to an AI that you have a definition of a term, which could be useful in an answer. In Chapter 9 we discussed content hubs and topic clustering; from a technical angle, implementing those via internal links and taxonomy is key so that crawlers can see the full picture of your content network. One emerging idea, relevant to both crawlability and AI, is the proposal of an llms.txt file as a companion to robots.txt . Introduced in late 2024, llms.txt is a concept by which site owners would create a special file to guide large language models to the most important information on the site ( ^[11] ) ( ^[12] ). Unlike a sitemap (which lists all pages for search indexing), an llms.txt would provide a curated, markdown-formatted overview of the site’s content specifically for AI consumption. The rationale is that LLMs have limited context windows and struggle to parse complex web layouts; a concise markdown guide can point the AI to key pages or provide summaries. For instance, llms.txt might include a brief description of your site and direct links to your documentation, FAQs, product pages, etc., in a simplified format. This standard is still in proposal stage, not widely adopted yet, but it signals how the industry is thinking about making websites more LLM-friendly at the source. Forward-thinking organizations (especially those with extensive documentation or data) may consider experimenting with such a file to see if it improves how AI agents interact with their content. At minimum, staying aware of initiatives like llms.txt will prepare you to take advantage once search engines or AI tools start looking for it. In summary, don’t lock your content behind technical barriers. Let legitimate bots in, and give them a map (sitemaps, internal links, perhaps future LLM guides) to roam your site easily. This foundational step is crucial: no amount of content optimization will matter if bots can’t access your pages in the first place. By being crawl-friendly , you ensure your carefully crafted content is visible to both the index and the algorithms that generate rich answers on top of that index.

10.2 Structured Data and Schema Markup

Structured data (schema markup) is a technical SEO cornerstone that has gained renewed significance in the era of generative AI search. By adding structured data to your HTML, you provide explicit clues about the meaning of your content – clues that search engines and AI can easily parse. In traditional SEO, schema markup has been used to enable rich results (like star ratings, recipe info, FAQ dropdowns in Google’s SERPs). Now, those same structured annotations can help your content become the building blocks of AI-generated answers and overviews.

10.2.1 Speaking the AI’s Language with Schema

Think of schema markup as a way of translating your human-friendly content into a format that machines can understand unambiguously. For example, if you have a product page, adding Product schema (with fields for name, price, availability, aggregate rating, etc.) tells search engines exactly what the key attributes of the product are. If you have an article, Article schema can specify the headline, author, publish date, etc. This metadata gives context that might not be immediately obvious from the raw text. Google has stated plainly: “You can help us by providing explicit clues about the meaning of a page by including structured data on the page.” ( ^[13] ) In one Google example, they mention that on a recipe page, schema can highlight the ingredients, cooking time, temperature, calories, and so on ( ^[13] ). In short, schema helps ensure that what your page is about is crystal clear to a machine. During 2023–2024, Google continued to invest in structured data. Notably, in early 2024 Google added support for product variant schema (to better understand product options) and introduced documentation for structured data carousels that appear within SGE results ( ^[14] ). These moves indicate that Google’s generative AI overview is utilizing schema.org data where available. In fact, SEO experiments have found that pages with comprehensive schema are more likely to be trusted and used by Google’s AI. As one agency noted, “A properly marked up site helps you to appear in AI answers by telling search engines what your data means (not just what it says) so they can accurately interpret the content.” ( ^[15] ) Structured data essentially acts like a fact highlighter, potentially making it easier for AI systems to extract relevant details or to choose your page as a source. Beyond Google, other AI-driven platforms also appreciate structured info. Bing’s AI chat, for instance, often cites sources and could benefit from schema to identify specific answers (like FAQs or how-to steps). Likewise, tools like Perplexity AI – which provides citation-rich answers – might more readily surface a site that clearly marks up Q&A or other useful content chunks. Even if an AI doesn’t explicitly read the JSON-LD on your page, remember that the search indexer does , and the AI often works off the search index. So better understanding by the indexer can translate to better inclusion by the AI.

10.2.2 Types of Schema to Implement

There are hundreds of schema types, but you don’t need to use them all – focus on those most relevant to your content. Here are some high-impact schema types for typical sites and how they help: Organization : Mark up your organization’s details (name, logo, contact info, social links). This reinforces your brand identity to search engines ( ^[16] ). It’s especially useful if an AI is answering a query about your company or needs to pull your logo or address for an overview. Breadcrumb : Provides the page’s position in your site hierarchy (e.g. Home > Category > Subpage) ( ^[17] ). This helps search/AIs understand site structure and can be used to display breadcrumb navigation in results. Article/BlogPosting : For content pages, this defines the headline, author, publish date, article body, etc. ( ^[18] ). In an AI context, clearly indicating the author and date can lend credibility (e.g. SGE might show the date to users). It also ties into Google’s emphasis on experience/expertise (E-E-A-T) by linking content to author entities. Product : Critical for e-commerce, this schema defines product name, description, price, currency, availability, reviews, etc. ( ^[18] ). If someone asks an AI “What’s the price of [Product]?” or “Is [Product] in stock?”, an AI overview might draw on this info. Indeed, Google’s AI snapshots have been seen displaying product specs and images, likely informed by structured data. FAQPage : Mark up frequently asked questions and answers on your page ( ^[19] ). This one is very powerful – many sites have an FAQ section, and marking it up can make you eligible for FAQ rich results. Moreover, an LLM can easily use a Q&A pair from your markup to directly answer a user’s question (with attribution). If ChatGPT or Bard is asked a question that exactly matches one of your FAQs, there’s a chance your Q&A could be used verbatim if the AI has access to that information. HowTo : If your content explains how to do something in steps, use HowTo schema to mark the steps, tools required, etc. Google’s SGE has shown step-by-step answers for how-to queries, often sourced from well-structured how-to pages. The HowTo schema makes it straightforward for an AI to identify the ordered steps on your page. LocalBusiness : For businesses with physical locations, this schema can provide your address, opening hours, geo-coordinates, etc. ( ^[20] ). An AI assistant answering “Find me a hardware store open now” could rely on such info. Person / Author Profile : Use Person schema for your authors or notable individuals on your site ( ^[20] ). This can reinforce expertise by linking content to author profiles (with details like their title, bio, sameAs links to social media). Google’s guidance around E-E-A-T suggests that clearly identifying authors and their credentials can improve content trust – something especially relevant if AI is summarizing advice or info from your page. And the list goes on – Recipe schema for recipe sites, Review schema for review content, Event schema for event listings, etc. The key is to identify the schema that aligns with your content and implement it consistently . Do an audit of your site: if you have a bunch of pages that could be marked up as FAQs, do it; if you have product pages without Product schema, add it. Each piece of structured data is another hint to the algorithms about what your page offers. To illustrate, here’s a small example of FAQ schema in JSON-LD (a common format for adding schema): html In the above snippet, we clearly label a question and its answer. A search engine or AI parsing this knows exactly that this text is a question (“How do I optimize my site for AI search results?”) and the answer to that question. In a scenario where, say, Google’s Bard is compiling an answer about “how to optimize for AI search,” it might pick up this Q&A from a site if it finds it relevant, thanks to the precise labeling. (Of course, having this markup doesn’t guarantee you’ll be featured, but it puts you in the game.) Structured data also contributes to your site’s authority and trust in the eyes of algorithms. A well-marked up site tends to be seen as well-maintained and transparent. As an SEO expert pointed out, “A well marked-up site is more trusted by Google than a poorly marked-up site. This is because Google can quickly and easily verify your site if you provide links to external reviews, social channels etc. through schema.” ( ^[21] ). In other words, adding schema like Organization with your social media profiles, or Product schema with real review data, helps connect the dots for Google’s knowledge graph. This can only help your content’s chances of being selected by an AI summary that values authoritative, well-sourced information.

10.2.3 Schema Markup Best Practices in 2025

When implementing schema, follow these best practices to get the most benefit: Ensure accuracy and consistency : The structured data must match what’s on the visible page. Don’t mark up a product price of $19.99 if your page says $24.99, for example. Inconsistent or erroneous schema can backfire (search engines might ignore it or even penalize gross discrepancies). Google explicitly requires that schema reflect the page’s actual content ( ^[22] ). Use JSON-LD format when possible : JSON-LD is Google’s recommended format because it’s easy to add without altering HTML elements. It goes in the <head> or anywhere in the HTML. Other formats like microdata or RDFa also work but can be messier to implement. Validate your markup : Use Google’s Rich Results Test or Schema.org’s validator to check that your JSON-LD has no syntax errors and is pulling the intended values from your page. Also, monitor Google Search Console for any Structured Data errors or warnings. Stay up-to-date with schema types : Schema.org periodically updates with new types/properties (e.g. the Product variant expansion). Keep an eye on SEO news or Google’s announcements for new supported schema that might give you an edge. For instance, if you run an e-commerce site and Google starts supporting a new ShippingDetails schema for AI shopping results, you’d want to implement that sooner than later. Prioritize high-impact pages : If you have a huge site, adding schema everywhere can be daunting. Focus on pages that drive your business goals and are likely to be used in AI answers. Typically, these are informational pages (for question answering) and key product/service pages. You can gradually expand coverage, but make sure the most important content is marked up first. Leverage schema for multimedia : Generative search is not just about text – Google’s AI Overviews can include images and videos. You can use schema to provide context for media too. For example, ImageObject schema can describe an image (caption, license, creator), and VideoObject can do similar for videos. This metadata could become more important as AI results get more visual (see Chapter 14 on multimodal search). In essence, adding structured data is like creating an enhanced resume for your content – it highlights all the key points in a way a machine can quickly grasp. As AI continues to evolve, feeding it structured, unambiguous information will only become more beneficial. Make your site scream its meaning, and you increase the odds that an AI will pick up on your content and present it to users in rich new ways.

10.3 Semantic HTML and Readability

While structured data is one layer of optimization, the very structure of your HTML content – the headings, paragraphs, lists, and other elements you use – also plays a huge role in how AI systems interpret and excerpt your site. Semantic HTML refers to using HTML elements according to their meaning (e.g. <h1> for the main title, <h2> for subsections, <article> for a standalone content unit, <ul> for a list, etc.), rather than just for visual formatting. Clean, semantic HTML combined with a clear writing style can make your content “dense with meaning” and easy for large language models to digest.

10.3.1 Structure is as Important as Keywords

In the old days of SEO, one might focus on sprinkling keywords in the text. In the AI era, it’s more about structuring your information logically . Large language models don’t simply scan for keywords; they ingest the page and build an understanding from the sequence of words and how the content is organized ( ^[23] ). One SEO expert explains that LLMs like GPT-4 or Google’s Gemini examine things like “the order in which information is presented, the hierarchy of concepts (which is why headings still matter), formatting cues like bullet points, tables, [and] bolded summaries.” ( ^[24] ). In other words, the model pays attention to your content’s outline and emphasis to figure out what’s important. If your page is a jumbled wall of text, an AI might struggle to find a clear answer or may misinterpret which parts of the text are most crucial. Consider an AI summarizing a lengthy article. How does it decide what the key points are? Likely, it will give extra weight to text that is prominent or structurally significant : titles, headings, list items, the opening sentences of a paragraph (which often contain the topic sentence), etc. If you use headings and subheadings effectively, you’re essentially giving the AI a mini road-map of your content. For example, an <h2>Benefits of Solar Panels</h2> followed by a concise paragraph and a bulleted list of benefits is far more accessible to an AI (and a human) than a page of unstructured paragraphs burying those benefits in fluff. In fact, well-structured content can outrank or outperform a keyword-stuffed page in AI results; “poorly structured content – even if it’s keyword-rich and marked up with schema – can fail to show up in AI summaries, while a clear, well-formatted blog post without a single line of JSON-LD might get cited or paraphrased.” ( ^[25] ). This underscores that content architecture is king. Schema is helpful but cannot compensate for a lack of clarity in the content itself. To optimize for this, write and format your content with both readers and AI in mind: Use a logical heading hierarchy : There should ideally be one <h1> (the page title), and then <h2> for main sections, <h3> for subsections, and so on. Each section should stick to a single topic or idea, as if you were writing an outline. This not only helps readers scan, but ensures an AI summary can pick out the section relevant to a particular question. For instance, if someone asks “What is the process to do X?” and your article has a section <h2>How to Do X: Step-by-Step</h2> , an AI can jump straight to those steps. Write descriptive headings : Instead of clever puns or vague headings, be straightforward. A heading like “10.3 Semantic HTML and Readability” (like we used above) is clear about the topic. If this were a vague heading like “The Secret Sauce,” an AI might not glean what that section covers until parsing all the text. Descriptive headings (potentially with relevant keywords) improve comprehension for models and humans alike ( ^[26] ). Keep paragraphs and sentences concise : Long, run-on paragraphs can dilute meaning. Aim for paragraphs that convey one idea and aren’t overly long (roughly 3-5 sentences each is a good rule of thumb). This creates natural pausing points and makes it easier for an LLM to extract a self-contained nugget of information from a paragraph without needing excessive context. Notice how in this chapter, most paragraphs are reasonably short – this is intentional for readability and excerptability. Use bullet points and numbered lists wherever appropriate. Lists are fantastic for both visual and algorithmic consumption. They break complex information into digestible chunks. For AI, a list clearly indicates a set of related points or steps. If a user asks “What are the main features of product Y?” and your page has a bullet list of features, an AI can easily turn that into a concise answer. Google’s generative search often presents answers in list form when the source content is in a list. In our experience, “Google AI Overview prefers well-structured, skimmable content. Ensure your articles include clear headings, short paragraphs, bullet points, [and] numbered lists for quick scanning.” ( ^[27] ). Highlight key information : If you have critical facts, definitions, or takeaways, consider using bold or italics to make them stand out (sparingly, when truly warranted). An AI model might notice emphasis. Similarly, a short summary sentence in bold at the start of a section (sometimes called a TL;DR or key point) can telegraph the main idea. Some websites put an important conclusion in bold text – which could be the line an AI chooses to quote directly. Use tables for structured data comparisons : When you have data or a comparison that fits a table format, using an HTML <table> can be helpful. Tables explicitly organize information into rows and columns. For instance, a pricing comparison table or a specs comparison (Feature X vs Feature Y) could be read by AI to pull a specific comparison point. (Do ensure to include a summary in text as well, since extremely tabular data might be skipped by some models that focus on text.) Include alt text for images (briefly, as a semantic point): While images themselves aren’t directly “readable” by text-based LLMs, the alt text you provide is. And with multimodal models emerging, having descriptive alt text ensures the AI knows what an image contains. For example, if you have a chart showing data, an AI like Bing’s image interpretation might read the alt text/caption to understand it. Below is a simplified example of well-structured HTML content: html <article> <h1>Guide to Solar Panel Installation</h1> <p>Installing solar panels can significantly reduce your energy costs. This guide outlines the steps and important considerations.</p> <h2>Benefits of Solar Panels</h2> <p>Solar panels offer multiple benefits for homeowners:</p> <ul> <li><strong>Lower electricity bills:</strong> Generate your own power and rely less on the grid.</li> <li><strong>Environmental impact:</strong> Solar energy is renewable and clean, reducing your carbon footprint.</li> <li><strong>Increased home value:</strong> Homes with solar installations often appraise higher.</li> </ul> <h2>How to Install Solar Panels</h2> <p>Here is a step-by-step overview of the installation process:</p> <ol> <li><strong>Assess your roof:</strong> Ensure it has structural integrity and good sun exposure.</li> <li><strong>Choose a system:</strong> Select solar panel type and inverter based on your energy needs.</li> <li><strong>Hire a professional (recommended):</strong> A certified installer will mount panels and connect the system safely.</li> <li><strong>Inspection and connection:</strong> Get the system inspected and connected to the grid per local regulations.</li> </ol> <h3>Common Mistakes to Avoid</h3> <p>Be aware of these pitfalls during installation:</p> <ul> <li>Not checking local permits and regulations.</li> <li>Ignoring the angle and direction of panels (affects efficiency).</li> <li>Skimping on quality for cost – cheaper panels may underperform long-term.</li> </ul> </article> In this snippet, the content is organized with meaningful headings (“Benefits of Solar Panels”, “How to Install Solar Panels”, “Common Mistakes to Avoid”). Important phrases are bolded to draw attention. Lists are used for benefits, steps, and mistakes, breaking the info into clear points. An LLM reading this would have an easy time identifying, say, the benefits of solar panels if asked, or enumerating the installation steps, because the HTML layout itself delineates those pieces. This is far better than a single giant paragraph about installation buried somewhere.

10.3.2 Write in a Clear, Conversational Style

Semantic HTML deals with the structure; equally important is the language style you use. Generative AI is essentially trying to emulate human answers. If your content is written in a plain, conversational manner that directly addresses common questions , it’s more likely to be selected and reproduced by an answer engine. A few tips on style and clarity: Address likely user questions in the text . Chapter 12 goes into prompt optimization, but from a writing perspective, it helps to pose and answer questions within your content. For example, include an explicit question as a subheading (“How much money can solar panels save annually?”) followed by the answer. This Q&A style content (even outside of an FAQ section) makes it trivial for an LLM to match a user’s question to your answer. In contrast, if the answer is hidden in a long narrative, it might be overlooked. Use natural language and define jargon . Content that reads in a straightforward way will be more quotable by AI. If you must use technical terms, define them briefly – not only is that good for users, it also helps the AI not to misinterpret specialized terms. Avoid unnecessary fluff . While a human reader might appreciate a bit of storytelling, an AI summarizer is looking for facts and direct statements to extract. It’s fine to have a personable tone, but try not to bury key facts in metaphor or overly flowery language. A generative AI might miss the nuance or, worse, mis-summarize it. Ensure each paragraph has a topic sentence . A well-crafted first sentence of a paragraph that summarizes the point acts as a signal to an AI. If the rest gets truncated, at least the main idea was clear up front. Maintain context : LLMs have limits on how much of the page they can use at once. If your page is very long, consider breaking it into sections or pages (perhaps with jump links) so that each addresses a subtopic clearly. Multi-turn AI conversations (like in Bing’s chat mode or others) might drill down into subtopics – if your content is modular, it fits these follow-up questions well. Use examples or analogies carefully : These can clarify for humans, but ensure you explicitly state the point the example is illustrating. An AI might otherwise repeat the example literally without the context, which could be odd. (For instance, if you say “Think of schema as the DNA of your site…” in an article about schema, Bard might respond with that analogy verbatim to a user – which may or may not be the ideal answer.) In summary, think of your page as an outline of answers . The better organized and clearer it is, the easier you make an AI assistant’s job. This not only improves your chances of being featured, but also reduces the risk of an AI misinterpreting or misrepresenting your content. By using semantic HTML and a reader-friendly writing style, you essentially future-proof your content for both human readers and AI algorithms that thrive on clarity and structure ( ^[24] ) ( ^[25] ).

10.4 Page Experience and Performance

No matter how great your content and markup are, a poor user experience can undermine it all. Page experience – which includes factors like site speed, mobile-friendliness, security, and lack of intrusive interstitials – remains a priority in the generative era. Google has repeatedly affirmed that the same signals used in regular search ranking continue to apply for AI features ( ^[28] ) ( ^[29] ). Fast, smooth websites not only rank better; they also integrate more seamlessly with AI systems that fetch and display content. In this section, we’ll look at why performance and UX still matter for GEO (Generative Engine Optimization) and how to ensure your site meets modern standards.

10.4.1 Core Web Vitals and Speed: Still Critical

Google’s Core Web Vitals – a set of metrics for loading performance, interactivity, and visual stability – are essentially a quantified measure of user experience. They include Largest Contentful Paint (LCP), Cumulative Layout Shift (CLS), and (recently replacing First Input Delay) Interaction to Next Paint (INP). As of 2025, Google treats these vitals as a key element of its page experience criteria ( ^[30] ) ( ^[31] ). In plain terms, Google wants websites to load quickly, not jump around as they load, and respond promptly to user input. Sites that meet the thresholds for “good” Core Web Vitals are likely to have a minor ranking advantage, and more importantly, they keep users happy. Why does this matter for generative search? Several reasons: User Behavior : Imagine a user sees an AI-generated answer with a source link to your site. If they click your link and your page loads slowly or is clunky, the user might bounce quickly. Not only have you lost that engagement (and potential conversion), but if this happens frequently, it could indirectly signal to Google that your page isn’t a satisfying result. In the context of SGE, Google wants to send users to helpful sites. It stands to reason that a fast site with a good experience is more likely to be deemed “helpful” than a sluggish site, all else being equal (even if just through correlation of other ranking signals like bounce rate or time on site). AI Content Fetching : Some AI agents fetch page content in real-time when generating answers (for example, Bing’s chat mode will visit webpages to quote them, and tools like Perplexity load pages to pull facts). If your site is extremely slow or has aggressive anti-bot measures, the AI might time out or fail to retrieve the info. A fast-loading site ensures that when an AI or bot pings your page, it can quickly get the content and move on. One can imagine that if Bing’s crawler encounters timeouts on your site, it might avoid using it as a source in the future due to reliability issues. Mobile-First Users : The majority of searches are on mobile devices, which often have slower connections. A page that loads fast on mobile (and is mobile-friendly) is going to serve those users better when they click through from an AI result on their phone. If an AI result encourages a user to visit your page, you want the transition to be frictionless. Google’s page experience guidelines explicitly emphasize mobile responsiveness and performance ( ^[32] ) ( ^[33] ). Future AI Integration : As AI features might become more directly integrated into browsers or assistant devices, having a lightweight page can facilitate quick previews or snippet generation. For example, if a voice assistant of the future fetches your page to read an answer aloud, you’d want it to fetch, parse, and start delivering that answer near-instantly. So, speed matters. Let’s drive the point home with some stats. Users are impatient: 53% of people will leave a mobile page if it takes longer than 3 seconds to load ( ^[34] ). Furthermore, fast sites have a clear business advantage. One study found that for B2B websites, a site that loads in 1 second had a conversion rate 3 times higher than a site that loads in 5 seconds (and 5 times higher than a site that loads in 10 seconds) ( ^[35] ). The relationship between load time and conversions/bounce is dramatic. We can visualize this: Figure 10.1: Impact of page load time on conversion rates. Faster pages yield significantly higher conversions – a 1-second load can see ~3× the conversion rate of a 5-second load, based on B2B website data ( ^[35] ). As shown above, users reward speed. Every additional second of loading can sharply reduce the likelihood of engagement or purchase. This is why Google continues to underline performance. In 2025’s page experience update, Google’s checklist for webmasters is to “perform well on Core Web Vitals” and keep improving speed via techniques like optimized scripts or server-side rendering ( ^[30] ) ( ^[36] ). In short, speed is a feature , not just a technical detail. To ensure your site meets these standards: Measure your Core Web Vitals using tools like Google PageSpeed Insights, Lighthouse, or the Core Web Vitals report in Search Console. Identify if LCP, CLS, or INP are in the “needs improvement” or “poor” range on either mobile or desktop. Optimize your assets : Compress images (they are often the biggest contributors to slow LCP), use modern image formats (WebP/AVIF), and serve images at appropriate sizes. Minify and combine CSS/JS files where possible, and defer loading of any scripts not needed for initial paint. Leverage browser caching and CDNs to reduce repeat load times and serve content from geographically closer servers. Use performance-enhancing techniques : such as lazy-loading images (via the loading="lazy" attribute on <img> tags for below-the-fold images), preloading critical resources, and removing render-blocking resources. For example, adding loading="lazy" in an image tag will delay loading that image until it’s needed, speeding up initial render: html <img src="/images/large-diagram.png" alt="Architecture Diagram" loading="lazy"> Mobile-first design : Ensure your responsive design is efficient. Avoid huge CSS frameworks or heavy libraries if not needed. Test on real devices or emulators to see how your site performs on a typical 4G connection. Avoid heavy client-side rendering for basic content : If your content is primarily text and images (like a blog), you likely don’t need a massive single-page app framework. Server-rendered HTML is fast and SEO-friendly. If you do use client-side frameworks, use dynamic hydration or static generation to send down HTML first (so the user isn’t staring at a blank screen). Monitor and iterate : Performance optimization is ongoing. Use real-user monitoring (e.g. Chrome User Experience Report data accessible via tools) to see how changes impact actual users over time.

10.4.2 Other Page Experience Factors

Performance is a big chunk of page experience, but not the only part. Google’s page experience update (and common sense) include several other factors: Mobile-Friendly, Responsive Design : As noted, your site should work well on mobile devices. This isn’t optional – Google moved to mobile-first indexing years ago. Responsive design (using CSS media queries to adapt layout) is the preferred approach ( ^[33] ). Test your pages in different screen sizes. Text should be readable without zooming, buttons/tap targets should be easily clickable, and horizontal scrolling should be avoided. If your desktop site is great but the mobile view is broken or hard to use, not only will users leave, but Google’s ranking for mobile searches (which feed SGE on mobile) will suffer. HTTPS Security : Serving your site over HTTPS is a must (and has been a lightweight ranking factor for a long time). In 2025, Google treats HTTPS as table stakes – it won’t boost you just for being HTTPS, but not having it could hurt trust and rankings ( ^[37] ). Also, AI scrapers likely skip non-HTTPS sites or could flag them as less trustworthy. Always redirect HTTP to HTTPS, and consider HSTS to enforce it. Avoid Intrusive Interstitials/Pop-ups : If your content is hidden behind a giant popup (like a newsletter sign-up or an app install banner), it frustrates users. Google has guidelines against intrusive interstitials, especially ones that cover content on page load. For AI, think about it this way: an AI trying to read your page might get stuck or read the wrong text if a popup dominates the HTML. Even if the AI can bypass it, a user clicking through won’t be happy to find they have to close a modal to see the info promised. Keep any required interstitials small, or delayed, or better yet, use subtle banners. Google explicitly says to avoid overlays that take up too much screen, especially above-the-fold ( ^[38] ) ( ^[39] ). This includes things like cookie consent banners – try to use minimal ones that don’t block content (or utilize the browser’s built-in mechanisms where possible). Ad Experience : Sites overloaded with ads, especially at the top of the page, create a bad user experience. Google’s “page layout algorithm” and subsequent guidance penalize sites that shove content far below ads ( ^[38] ) ( ^[40] ). If a user comes from an AI answer expecting to see the solution and instead they get a full-screen ad or five ads before the content, they’ll bounce. Also, an AI summarizer might inadvertently read ad code or irrelevant text if the page isn’t well-structured. Keep ads to reasonable levels, and make sure they’re labeled and separated in the DOM (so, for instance, use dedicated containers or iframes that an AI can skip over as not part of main content). Consistent Layout (No Jank) : This relates to CLS (layout shift). Ensure that your CSS and media dimensions are set so that the page doesn’t jump around as it loads. Unexpected shifts can not only annoy users but might confuse an AI trying to capture a screenshot or parse content during load. Enable Prompt Content Display : When a user clicks through from an AI, they often have a specific query in mind (maybe even a specific snippet that was referenced). Consider using techniques like fragment URLs or highlighting. For instance, some search features scroll the user to the quoted text or highlight it (SGE was experimenting with this). You can’t fully control that, but having clear anchor links for sections could allow a browser or AI agent to jump to the relevant part (for example, a table of contents with anchor links to sections). Monitor with Real Users : Keep an eye on your analytics for bounce rates, time on site, etc., especially for traffic coming from new AI features. If you see unusual behavior (like very short time-on-page for AI-originating clicks), it might hint that users aren’t finding what they expected (or page load issues). This feedback loop can inform further UX tweaks. The bottom line is that user experience principles haven’t changed – if anything, they’re reinforced . Google’s own documentation ties helpful content with good page experience, noting they are “fundamentally connected” ( ^[41] ). The companies building AI search want to provide good experiences, so they will naturally favor content from sites that do the same. A great technical SEO knows that speed and UX improvements not only boost SEO, but also conversion and user satisfaction. It’s truly a win-win-win for users, search engines, and your business metrics. By investing in page experience, you ensure that when your content is surfaced – either directly in an AI answer or via a link – the user’s journey doesn’t falter. They get the information faster, they trust your site more, and they’re likelier to stick around. In a world where attention is gold, a snappy, pleasant website is your chance to shine after earning that AI-generated click.

10.5 Preventing AI Misinterpretation and Misuse

A unique challenge with generative AI is that it might reinterpret or repurpose your content in ways you didn’t intend. While traditional SEO is mostly about getting indexed and ranked, GEO also involves guiding how AI systems use your content . This includes preventing snippets from being taken out of context, avoiding hallucinations (where the AI might mix up facts), and generally ensuring your content is represented accurately. In this section, we’ll discuss technical measures to control or influence how AI “reads” your pages – from special meta tags that block AI summaries to using clear markup for quotes or code to reduce misinterpretation. We’ll also touch on emerging standards and ethical considerations for content usage.

10.5.1 Marking Content for Clarity (Quotes, Code, and More)

One straightforward way to avoid AI misinterpretation is to clearly delimit different types of content on your pages. By using the appropriate HTML elements for quotes, code, definitions, etc., you give the AI parser cues about the nature of that text. This can prevent, for example, a user comment or a sarcastic statement from being read as the site’s official stance. Consider a scenario: You run a forum or a Q&A site. A user posts an incorrect answer or a controversial opinion, and your page displays it. If that user content isn’t distinguished in markup, a search AI might scrape the page and present the user’s statement as a fact attributed to your site. That could be damaging or just inaccurate. By wrapping such content in a <blockquote> with a citation of the user, or marking it as a user-generated section, you at least signal “this is a quote/opinion”. Google’s indexing system might treat it differently (for instance, Google often ignores or devalues text in <blockquote> for snippet purposes if it’s clearly a quote from elsewhere). Likewise, an LLM might be more likely to attribute the quote properly or skip it if not relevant to a direct question. Similarly, for technical content, using the <code> or <pre> tags for code snippets or command-line outputs is critical. Not only does this preserve formatting, but it tells any AI or parser that “this text is code or technical output”. The AI then is less likely to confuse it with prose. For example, if you have a line in your tutorial that shows an error message or a piece of JSON, putting that in a code block ensures the AI doesn’t accidentally mingle it with your explanatory text.

It might also choose to display it verbatim (with a monospace font) if providing an answer. For instance: html<p>When we ran the test, we encountered the following error:</p> <pre><code>ERROR 503: Service Unavailable</code></pre> <p><em>Solution:</em> This error usually means the API endpoint is down; try again later.</p> In the above snippet, the error message is clearly marked as code. An AI summarizing common errors would likely quote the error exactly as shown (which is what you want), and it knows the next paragraph is a solution (since it’s in normal text with perhaps emphasis on “Solution:”). By contrast, if you had just written: “When we ran the test, we encountered ERROR 503 Service Unavailable solution: this means the API is down…”, the AI might extract something garbled. Another use of markup is for definitions or key terms. You might use <dfn> tag for defining instances (though not widely used) or simply italic/bold the first occurrence of a term and define it immediately. For example: “Generative Engine Optimization (GEO) – adapting SEO techniques for AI-driven search results.” This immediately pairs the term with its definition. If someone asks an AI “What is Generative Engine Optimization?”, there’s a tidy definition it can pull from your page. Some advanced HTML5 elements like <aside> can mark side content, and <figcaption> can label image captions – use these appropriately so that if an AI scrapes your page, it can distinguish main content from side notes and captions. In summary, use HTML as it was intended, to semantically separate content roles. Quotes for quotes, lists for lists, headings for titles, code for code, etc. A well-structured HTML document not only looks organized, but semantically it minimizes misreads. It’s like giving AI a script with stage directions.

The model might still mess up, but you’ve done your part to clarify who is saying what and in what format.

10.5.2 Controlling Snippets and AI Usage via Meta Tags

While semantic HTML helps with interpretation, there are cases where you may not want your content to appear in AI-generated snippets at all, or you want to limit how much of it appears. Perhaps you run a subscription-based site and prefer not to have AI giving away your content for free, or maybe you have an page that you feel is likely to be misused if taken out-of-context. Google has provided some tools – originally for controlling search snippets – that also apply to its generative AI snippets in Search. By using these, you can opt out or limit how your content is used in AI overviews . Key methods (for Google in particular) include the following ( ^[42] ) ( ^[43] ): nosnippet meta tag – This tells Google not to show any snippet of your page in search results. Implement by adding to your HTML <head> : <meta name="robots" content="nosnippet"> . Google has confirmed this will prevent your content from being used in SGE AI overviews or featured snippets ( ^[42] ). Essentially, your page can still be indexed and ranked, but Google will only show the URL/title (no text extract). This is a blunt but effective tool if you absolutely want to avoid being summarized by Google’s AI. Keep in mind, it also removes your rich snippet in regular search, which might reduce clicks. Use it selectively. max-snippet meta tag – This meta directive lets you specify a maximum character length for snippets. For example: <meta name="robots" content="max-snippet: 50"> would tell Google to only use up to 50 characters of a snippet ( ^[44] ). Setting it to 0 is effectively the same as nosnippet (no snippet at all) ( ^[44] ). This gives a bit more nuance – you could allow a short snippet but not a long excerpt. Maybe you’re okay with a one-liner appearing in AI, but not a full paragraph. data-nosnippet attribute – This is an HTML attribute you can apply to specific elements in the body of your page to mark them as off-limits for snippets ( ^[45] ). For instance, <p data-nosnippet>Confidential information here.</p> ensures that particular paragraph won’t show up in Google’s results or AI answers ( ^[45] ). This is useful if 95% of your page is fine to snippet, but there’s a sensitive part (like a key takeaway that you want people to click through for, or a segment that doesn’t make sense out of context). By sprinkling data-nosnippet on those parts, you control exactly what content could be lifted. X-Robots-Tag HTTP header – This is similar to the meta tags, but set at the server level. You can configure your server to send X-Robots-Tag: nosnippet in the HTTP headers for a page ( ^[46] ). It has the same effect as the meta tag. This is often used for non-HTML content (like PDFs) or if you prefer server config. For most, the meta tag approach is easier, but it’s good to know both exist. Canonical tags for duplicates – If you have duplicate or very similar content on multiple pages, canonicalization helps ensure Google (and by extension its AI) knows which is the primary source ( ^[47] ). This can prevent weird cases where perhaps an AI overview pulls from a duplicate page or shows a less complete version of your content. By using <link rel="canonical" href="https://www.example.com/preferred-page"> on duplicates ( ^[47] ), you signal the original. This is a standard SEO practice, but in the AI context it’s about steering the AI to the source you want. It also helps avoid confusion if, say, you have a print view of an article – you don’t want the AI quoting the print view URL. Let’s see a quick example of using some of these on a page: html <head> <meta name="robots" content="max-snippet: 0, noimageindex"> <!-- This would prevent text snippets and also avoid indexing images on the page --> </head> <body> <h1>Research Report on Industry X</h1> <p>Executive summary of the report...</p> <p data-nosnippet><strong>Key Findings:</strong> [The key findings are listed in the full report]</p> <p>The rest of the page content goes here...</p> </body> In this snippet, we chose to use max-snippet: 0 for the whole page (no text snippet at all) and also noimageindex just as an example to not index images (maybe if they were proprietary charts). Additionally, we put a data-nosnippet on the “Key Findings” paragraph. This belt-and-suspenders approach ensures the most crucial part (the findings) never show up in AI – forcing users to click the page to read them – and in fact no snippet at all will show. Alternatively, we could be less strict in the meta tag (allow some snippet) but still protect the findings paragraph. The combination is flexible. Important caveat : These measures currently apply mainly to Google Search and any of Google’s generative search features. Other platforms may not honor them. Bing, for instance, at one point said it would respect meta noindex and maybe nosnippet , but we don’t have as clear documentation on Bing Chat’s handling. That said, if you block Bing’s crawler via robots, it won’t see the content at all. OpenAI’s ChatGPT browsing plugin would obey robots (and thus not see pages disallowed). But if your content ended up in the training data of GPT-4 already, nosnippet now won’t retroactively remove it. So these controls are mostly about future AI interactions and specifically things like search engine generated answers. Pros and Cons : Using snippet controls is a double-edged sword. On one hand, it can drive more clicks (since users can’t get the info without visiting) and protect content. On the other, your site might not be referenced by AI at all if it can’t use a snippet. Google has noted that links in AI overviews often get higher CTR than traditional results ( ^[48] ) – if you opt out of being included, you miss out on that traffic. So use these tactics thoughtfully. Perhaps you employ them on pages where you genuinely need to withhold info, but leave most pages open for AI to feature. It’s analogous to the early days of featured snippets – some sites blocked them fearing loss of traffic, only to find they lost presence. Others embraced them and adjusted strategy (e.g., by providing just enough answer to entice a click for more detail). As a technical SEO, you should also monitor how your content appears in AI outputs . Search for your brand or content snippets on ChatGPT (with browsing or plugins), Bard, Bing, etc. If you find the AI is consistently misunderstanding or misusing your content, that might be a clue to tighten things up – maybe add data-nosnippet around the problematic bits, or add more clarifying text that eliminates ambiguity.

10.5.3 Emerging Standards and Keeping Control

The landscape of AI and content usage is evolving rapidly. We’ve seen the emergence of proposals like NoAI meta tags and the previously discussed llms.txt . While not yet standardized by any search engine, the “noai” directive is being promoted in some communities (especially among artists and content creators) as a way to signal “I don’t want my content used for AI” ( ^[49] ). For example, DeviantArt introduced a <meta name="robots" content="noai"> for art pages to opt out of AI training ( ^[50] ). Some platforms like Raptive (an ad network) have added support for noai in their publisher settings ( ^[49] ). It’s important to note that these are honor-system signals – currently, there’s no legal or technical enforcement making AI companies comply universally. OpenAI and Google’s approach (GPTBot and Google-Extended) is the more concrete opt-out for training. But we may see a broader adoption of a machine-readable “no AI usage” flag if regulations push that way. On the flip side, we might also see tags for allowing or specifically feeding AI. For instance, a hypothetical aisummary="allowed" attribute or a schema property that indicates a snippet is expressly license-free for use. The idea has been floated that publishers might label certain content as AI-summarizable. While not reality yet, being aware of these discussions means you can implement quickly if they become available. Another thing to watch is the regulatory environment. Governments are starting to discuss mandates around AI data usage transparency. It’s possible that in the near future, AI systems could be required to provide citations for all content or to exclude content that was disallowed. If that happens, technical SEO will include ensuring your preferences (to be included or not) are clearly communicated via whatever standard is decided (be it robots.txt, meta tags, or a new protocol). Monitoring tools : As part of your GEO efforts, consider tools or services that track your content’s presence in AI outputs. Some startups are emerging that claim to monitor if your website is mentioned or used by AI answers. Even simple Google Alerts or searches can catch when your text appears (though AI paraphrasing makes that tricky). If you find unauthorized or unwanted usage, you may decide to adjust your technical stance (e.g., start blocking a particular bot). For instance, if an obscure AI tool is scraping you too aggressively, you could block its user agent. Finally, educate and collaborate with your legal and content teams. Technical decisions like blocking AI bots or adding noai tags might have business implications. There’s a balance between protecting content and gaining exposure. Part of an SEO’s role now is to advise on that strategy. For example, a financial data provider might block AI to preserve their data’s value, whereas a blog seeking readership might welcome being referenced by AI for the added visibility. These aren’t just technical calls; they’re business calls enabled by technical measures. In conclusion, while you can’t perfectly control how AI will use your content, you do have some levers to pull. Use HTML semantics to avoid misunderstandings, and use meta directives to set boundaries with major AI-enabled platforms. Keep an eye on emerging standards like noai and llms.txt – even if they’re voluntary, they indicate a direction. By staying proactive, you protect your content’s integrity and ensure your SEO strategy adapts to the AI age rather than getting run over by it.

10.6 Conclusion and Key Takeaways

Technical SEO in the generative search era is all about laying a strong, adaptable foundation for your content. The core principles haven’t radically changed – you still need to be crawlable, fast, and structured – but the stakes are higher and the nuances are new. A quick recap of what we’ve covered in this chapter: Crawlability & Access: Make sure all the right doors are open. Let search engines and reputable AI crawlers index your content. Decide strategically on allowing or blocking crawlers like GPTBot and Google-Extended based on your comfort with AI training usage. Use tools like robots.txt to communicate your preferences ( ^[3] ) ( ^[4] ). And don’t forget internal links and sitemaps to guide bots through your site. Structured Data: Speak in schema wherever possible. By marking up content with FAQ, HowTo, Product, and other schemas, you make it easier for search and AI to understand and trust your pages ( ^[15] ) ( ^[51] ). Schema is your content’s metadata resume – the extra mile that could win you that featured snippet or SGE inclusion. Implement relevant schema types and keep them updated as new ones emerge (for example, keep an eye on any schema that specifically aids AI results or rich media in search). Semantic HTML & Content Structure: Structure beats stuffing. Use headings, lists, and clear formatting to make your points stand out ( ^[24] ). Think about how an AI (or a rushed reader) would scan your page. Make it easy for them to pick up the main ideas and answers. A well-structured page is future-proof – whether it’s Google’s crawler, a GPT model, or the next big AI, the logic of your content will shine through. Page Experience & Performance: Speed and UX are the unsung heroes of SEO and now GEO. Users and AI both prefer fast, user-friendly sites. Optimize those Core Web Vitals, be mobile-friendly, and avoid anything that annoys or slows down visitors ( ^[35] ) ( ^[34] ). This not only helps rankings but ensures you convert the traffic you get. If an AI cites you and the user clicks, that click is half the victory – a great page experience seals the deal. AI Control & Snippet Governance: You have tools to prevent or shape how your content appears in AI answers. Use them judiciously. Meta tags like nosnippet or attributes like data-nosnippet can keep sensitive info out of AI overviews ( ^[42] ) ( ^[45] ). Proper HTML markup (for quotes, code, etc.) can reduce misinterpretation. Stay informed on new developments like the noai directive ( ^[49] ) or llms.txt ( ^[11] ) – they’re hints of a more structured future where content creators can explicitly signal AI usage rights and guidance. For online marketing professionals, the takeaway is that technical SEO and content strategy are two sides of the same coin . High-quality, E-E-A-T-rich content (as discussed in earlier chapters) needs a technically sound platform to truly succeed in generative search. You want your content not just to exist, but to be understood correctly and delivered optimally by the new generation of search tools. As you advise companies and work on websites, instill a mindset of “AI-readiness” in development and SEO practices. That means: Keeping website infrastructure up-to-date (fast servers, latest security, modern frameworks optimized for SEO). Ensuring new content is published with schema and clear structure from the get-go (perhaps create templates that enforce this). Regularly auditing robots.txt and meta directives as the search landscape changes (maybe the defaults we use today will change if, say, a major search engine decides to use a different crawler or require an explicit opt-in for AI features). Coordinating with content creators to place important info in places (or formats) that will get noticed by AI. For example, if there’s a critical statistic or quote, maybe make it a one-sentence paragraph or a call-out that an AI won’t miss. Monitoring performance and logs – watch for any crawler issues, page speed regressions, or unusual bot activity that could indicate an AI scraping your site in undesirable ways. Embracing new standards early – being among the first to implement something like llms.txt could give an advantage if LLM-powered services start looking for it to enhance their answers. In essence, technical SEO for GEO is about being proactive and detail-oriented . Small tweaks (like a meta tag here, an alt text there, a 0.1s load improvement) can compound into a significant edge when multiplied across thousands of queries and users. It’s akin to tuning an engine – each adjustment might only improve things slightly, but together they make your site a high-performance machine in the race for AI-age visibility. The companies that master technical SEO in this era will find that their content reaches not only more people but the right people at the right moments – whether through a chatbot, a voice assistant, or the evolving search result pages. By following the strategies in this chapter, you’ll ensure the technical fidelity of your site matches the excellence of your content, creating a synergy that propels your online visibility to new heights, no matter how search evolves. Sources: Google Search Central – AI features and your website ( ^[29] ) ( ^[9] ) SearchEngineJournal – Shelby, C. How LLMs Interpret Content ( ^[24] ) ( ^[25] ) Edge45 – Walker, J. Optimising for AI Overviews using schema ( ^[14] ) ( ^[21] ) Cyberchimps – How to Rank in Google AI Overview ( ^[51] ) ( ^[26] ) Stan Ventures – Thekkethil, D. Keep Your Content Out of SGE Overviews ( ^[42] ) ( ^[45] ) EFF – Klosowski, T. Opt Out of ChatGPT and Bard Training ( ^[3] ) ( ^[4] ) The Verge – Roth, E. Fewer websites are blocking OpenAI’s crawler ( ^[6] ) AIbase – GPTBot Blocking Rate Increases ( ^[5] ) SiteBuilderReport – Website Speed Statistics (2025) ( ^[35] ) ( ^[34] )

References

^[1] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[2] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/google-extended-crawler-432636

^[3] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[4] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[5] www.aibase.com - Aibase.Com URL: https://www.aibase.com/news/1768

^[6] www.theverge.com - Theverge.Com URL: https://www.theverge.com/2024/10/7/24264184/fewer-websites-are-blocking-openais-web-crawler-now

^[7] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[8] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[9] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/docs/appearance/ai-features

^[10] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/docs/appearance/ai-features

^[11] Llmstxt.Org Article - Llmstxt.Org URL: https://llmstxt.org

^[12] Llmstxt.Org Article - Llmstxt.Org URL: https://llmstxt.org

^[13] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[14] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[15] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[16] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[17] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[18] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[19] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[20] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[21] Edge45.Co.Uk Article - Edge45.Co.Uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up

^[22] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/docs/appearance/ai-features

^[23] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[24] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[25] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-llms-interpret-content-structure-information-for-ai-search/544308

^[26] Cyberchimps.Com Article - Cyberchimps.Com URL: https://cyberchimps.com/blog/how-to-rank-in-google-ai-overview

^[27] Cyberchimps.Com Article - Cyberchimps.Com URL: https://cyberchimps.com/blog/how-to-rank-in-google-ai-overview

^[28] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/docs/appearance/ai-features

^[29] Developers.Google.Com Article - Developers.Google.Com URL: https://developers.google.com/search/docs/appearance/ai-features

^[30] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[31] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[32] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[33] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[34] www.sitebuilderreport.com - Sitebuilderreport.Com URL: https://www.sitebuilderreport.com/website-speed-statistics

^[35] www.sitebuilderreport.com - Sitebuilderreport.Com URL: https://www.sitebuilderreport.com/website-speed-statistics

^[36] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[37] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[38] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[39] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[40] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[41] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/page-experience-seo-448564

^[42] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[43] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[44] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[45] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[46] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[47] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[48] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/blog/ai-overview-prevent-content

^[49] Help.Raptive.Com Article - Help.Raptive.Com URL: https://help.raptive.com/hc/en-us/articles/13764527993755-NoAI-Meta-Tag-FAQs

^[50] www.foundationwebdev.com - Foundationwebdev.Com URL: https://www.foundationwebdev.com/2022/11/noai-noimageai-meta-tag-how-to-install

^[51] Cyberchimps.Com Article - Cyberchimps.Com URL: https://cyberchimps.com/blog/how-to-rank-in-google-ai-overview

11. Off-Page SEO and Brand Authority in the AI Era

The rise of generative AI in search means that off-page SEO – the signals and content about your brand beyond your own website – has never been more critical. In traditional SEO, tactics like link-building and PR established authority and boosted rankings. In the AI era, those same off-page efforts take on new dimensions. Large language models (LLMs) and AI-driven search tools draw on vast swaths of web content to formulate answers. They tend to favor information from well-known, authoritative sources , and they often repeat the narratives and opinions prevalent across the web. As a result, building your brand’s authority off-site doesn’t just help with Google rankings – it can directly influence whether and how your brand is mentioned in AI-generated answers. In this chapter, we explore how to bolster off-page signals and brand authority in ways that resonate with AI models and generative search experiences . From digital PR and community engagement to leveraging reviews and open data, we’ll examine strategies (with real examples from 2024–2025) to ensure your brand is both trusted by AI and prominent in AI-driven results . We’ll also discuss tools for monitoring your brand’s presence in ChatGPT, Google’s Search Generative Experience, and other AI platforms, so you can continually refine your off-page strategy. The future of search visibility will be about more than just blue links – it will be about being part of the conversations and content that intelligent systems use to answer user questions.

11.1 Digital PR for Authority in AI Search

Off-page SEO has long been about establishing authority : earning backlinks, media mentions, and references from credible third-party sites. In the era of AI search, this digital PR aspect is even more crucial. LLMs like ChatGPT or Google’s generative search AI don’t literally calculate PageRank, but they are heavily influenced by what content they ingest and deem trustworthy. Generally, LLMs favor content from authoritative domains – those that are widely cited, well-known, or associated with expertise ( ^[1] ). This means securing mentions in trusted publications , getting experts to cite or quote your work, and having a presence on high-authority sites can increase the likelihood that AI models regard your brand as worthy of inclusion in answers. The “Mentions” Currency: Unlike Google’s ranking algorithm which is built on links, AI models prioritize mentions and context in their training data ( ^[2] ). SEO expert Rand Fishkin describes it succinctly: “The currency of Google search was links… The currency of large language models is mentions (specifically, words that appear frequently near other words) across the training data.” ( ^[2] ) In practice, if your brand or product is frequently mentioned alongside important keywords or topics (especially on respected sites), an LLM is more likely to recall or include it when generating an answer about that topic. For example, if many articles and lists about “top project management tools” mention a particular software brand, an AI answer to “What’s a good project management tool?” might very well include that brand by default , due to the patterns learned from those sources. Earning Credible Mentions: The key, then, is to get your brand talked about in the right places . Traditional PR techniques – press releases, thought leadership articles, expert interviews – can lead to coverage in news sites, industry blogs, or research papers. These are precisely the kind of high-authority, factual sources that AI models are trained on. A mention or quote in a New York Times article, a citation in a university study, or inclusion in a “Top 10” list on a reputable blog not only boosts your human credibility but also means that when an AI combs through text to answer a question, your brand has a foothold in that knowledge base. For instance, when OpenAI’s GPT-4 was asked about the “best brands for small business marketing,” it produced a list of well-known software brands with citations from Wikipedia and a NerdWallet review article ( ^[3] ). The brands that appeared – Constant Contact, Mailchimp, HubSpot, ActiveCampaign , etc. – were all those with strong digital PR: they are frequently reviewed by third parties and discussed in authoritative contexts. In the ChatGPT-generated excerpt below, we can see that these brands are surfaced with sources like Wikipedia and NerdWallet highlighted: ChatGPT’s answer (April 2025) to “What are the best brands for small business marketing?” cites a NerdWallet affiliate review and Wikipedia as sources. The response lists only well-known brands in the marketing software space, reflecting a bias toward companies with significant off-page presence and coverage. Smaller or less-cited brands are absent, underscoring how AI answers gravitate to what authoritative sources have discussed ( ^[3] ). Inclusion in such lists doesn’t happen by accident – it’s often the result of successful outreach and PR . A tactical example: if you run a boutique CRM software company, you’d want to be mentioned in “Best CRM” roundups on sites like PC Magazine , TechRadar , or relevant industry blogs. Even if those mentions start as part of an earned media effort (e.g., pitching your product for review or contributing an article), the long-term benefit is that when an LLM later “reads” those articles during training or via real-time retrieval, it learns that your brand is associated with that category and carries credibility . As Rand Fishkin advises, to get your brand into AI answers, you might need to “make sure that our brand is mentioned [on] the places on the web” that discuss your topic, even if “that’s a PR process and a pitch process” – it is absolutely worthwhile ( ^[4] ) ( ^[5] ). Authoritative Domains and LLM Trust: Search engines have long used domain authority or E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals as a proxy for credibility. LLMs don’t have an explicit E-E-A-T score, but they implicitly reflect these signals by relying on authoritative text. For example, if your company’s research report is cited by Wikipedia or appears on a .edu site, an AI model will likely treat the information as more reliable. Indeed, marketers have observed that ChatGPT’s newer browsing or integrated versions heavily cite Wikipedia for company or product information ( ^[3] ). Being on Wikipedia (with a well-sourced page) thus becomes an off-page priority – it’s a seal of notability that not only helps with Google’s knowledge panels but also virtually guarantees an LLM like GPT knows about your brand. In April 2025, one test found that ChatGPT cited Wikipedia five different times when listing top marketing brands ( ^[3] ). While not every brand can have a Wikipedia page, you can aim to be referenced within Wikipedia articles (for instance, a study or infographic your company published might be cited as a source on a relevant Wikipedia page). Such references mean your brand’s name or URL is literally in the training data that many models ingest. Digital PR Case Study – HubSpot: A real-world example of digital PR’s impact involves HubSpot, a prominent marketing software company. HubSpot invests heavily in content and PR, resulting in numerous off-page mentions. When SEO experts queried a generative AI (ChatGPT-4) for “Tell me about HubSpot,” the answer came back with a well-rounded summary – and multiple citations to HubSpot’s own site and external sources ( ^[6] ). Interestingly, the AI cited HubSpot’s legal page and knowledge base, but not the official About Us page ( ^[7] ). This indicated that those were the pages where the model found the most concrete info about the company’s background (possibly because HubSpot’s about page was sparse on details). The takeaway for HubSpot’s team was to enrich their About page with more substantive content ( ^[8] ) – yet the bigger picture is that HubSpot’s widespread presence (press coverage, a large Wikipedia entry, many third-party articles) ensured the AI had plenty of material to draw from. Smaller brands with little third-party recognition often get no mention at all in similar AI queries. In other words, if the world at large isn’t talking about your brand, AI likely won’t either . Building digital PR and off-site authority is how you change that. Impact on AI Citations and Summaries: Strong off-page authority not only influences whether an AI mentions you, but how . Google’s Search Generative Experience (SGE) and Bing Chat both provide AI-generated summaries with citations. These citations tend to favor sites Google/Bing already trust (news outlets, popular blogs, high-authority domains). If you’ve executed a good PR strategy, your brand might be the source being cited in those AI overviews. For instance, a well-placed thought leadership piece by your CEO on TechCrunch could lead to SGE citing TechCrunch (and by extension mentioning your CEO or brand) in an overview on that topic. Even if your site isn’t the one cited, getting an influential publication to include and credit your insights means the AI can pick up on that content. In short, off-page authority creates a ripple effect in AI: it increases the odds of your brand being part of the AI’s knowledge, and it elevates the credibility of your own content in the eyes of AI. It’s also worth noting that LLMs have a popularity bias : they are more likely to mention well-known entities simply because those appear frequently in the training data ( ^[9] ) ( ^[10] ). A 2023 study on brand bias in LLMs found that models “favor established global brands while marginalizing local ones,” often associating big brands with more positive attributes ( ^[11] ) ( ^[10] ). This is a double-edged sword. On one hand, it means incumbent brands with lots of press and mentions enjoy an advantage – AI might default to recommending them. On the other hand, it’s a challenge for up-and-coming players. The only way to counteract that bias is to increase your share of voice in reputable sources . If you can generate a steady drumbeat of coverage – say, a mention in a Forbes article here, a quote in a Gartner report there – over time the AI’s “memory” of your brand grows. In fact, marketing strategists in 2025 often talk about “feeding the model”: ensuring that at least some of the billions of words being used to train GPT-4, GPT-5, Gemini, etc. include your brand in a positive, relevant light. Peter Buckley of Meta argued that “we now need to market to two brains: the human and the model” . Human audiences respond to emotional storytelling, but “LLMs rank brands based on facts, structure, and relevance” , meaning brands need factual, cite-worthy content in the public domain ( ^[12] ). Multiple studies, he notes, show that “the volume of brand mentions across the web directly affects whether a brand is surfaced by LLMs” , giving earned media a whole new level of strategic value ( ^[1] ). In other words, every PR win not only boosts brand awareness for people, but also plants a seed in the AI ecosystem that can yield visibility later. Tactical Tips for Digital PR in the AI Era: Aim for High-Authority Coverage: Prioritize getting features or mentions in sites that AI models are likely to “trust” (news sites, popular Q&A sites, .edu and .gov resources, etc.). A single reference in Wikipedia or a respected journal can do more for AI visibility than dozens of minor blog mentions. Create Original Research/Data: One effective PR tactic is publishing studies or reports that others will cite. If your data point gets picked up widely (e.g., “Company X’s survey finds 60% of consumers do Y”), it might end up referenced across the web and even in knowledge repositories like Wikipedia. Generative AI often regurgitates such statistics if they were prominent in its training set. Original, citable content thus serves both as link bait and as “AI bait.” (For example, the Bain & Company stat that “80% of consumers rely on AI content for at least 40% of their searches” is now echoed in many articles ( ^[13] ) – any AI trained on 2024 web data has likely absorbed that tidbit.) Leverage Thought Leadership: Get your experts (or brand) quoted on authoritative platforms. A quote in a New York Times article or a mention on a popular podcast that later gets transcribed can become part of the AI narrative about your industry. LLMs like to name-drop known experts and companies when answering questions like “what do experts say about X?” – so being among those quoted voices enhances your AI visibility as a recognized authority. Monitor and Amplify Earned Media: Whenever your brand does get a high-profile mention, make sure to amplify it (through social channels, press sections on your site, etc.) and possibly get it linked or referenced elsewhere. The more it spreads, the more entrenched it becomes in the training data supply. Consider also connecting with the author or site to correct any inaccuracies – you want the information that propagates to be accurate, since AI might later present it as fact. (This is part of reputation management in the AI context – an extension of PR where you not only get mentioned but ensure the mentions are correct.) In summary, digital PR and off-page authority tactics lay the groundwork for brand inclusion in AI-driven results . They establish your brand as one of the names that comes up in credible discussions – and thus one of the names an AI might confidently present to a user. As generative search grows, those brands that have invested in off-site authority will often find themselves one step ahead in the new zero-click, AI-answered world.

11.2 Social and Community Presence: Feeding the Conversational Web

In the modern search ecosystem, content isn’t confined to websites and news articles . A huge amount of knowledge and discussion lives on social networks, forums, and community-driven platforms. Users ask questions on Reddit and Quora, share reviews on YouTube and TikTok, and offer advice in specialist communities. Importantly, LLMs sometimes draw from these less-formal sources – either through their training data or via real-time search integration. Therefore, participating in online communities and cultivating a social media presence can indirectly influence what AI systems “know” and say about your brand or product. Forums and Q&A Sites: Platforms like Reddit, Stack Exchange, Quora, and niche forums are goldmines of crowd-sourced information. For years, Google’s search results have often surfaced Reddit threads or Quora answers for long-tail queries. Now, AI-driven engines are following suit. Tools like Perplexity.ai explicitly incorporate forum content in their cited answers. Google’s SGE has a “Discuss” or “Perspectives” feature that can pull in forum posts or social commentary for certain queries. Moreover, many LLMs were trained (up to their knowledge cutoff) on data from these platforms . Reddit, in particular, has been a significant part of AI training sets – so much so that in 2023 Reddit announced plans to charge for API access and data licensing to AI companies ( ^[14] ). The message is clear: popular subreddit discussions are valuable training data . If a topic or brand is frequently discussed on Reddit, an AI model might have “seen” those discussions. From a marketer’s perspective, this means engaging authentically in relevant online communities can pay dividends . Consider an example: A developer asks on Stack Overflow about the best cloud hosting for a certain use case. If an employee or advocate of a hosting company provides a helpful, non-promotional answer that gets upvoted, that Q&A might become part of the data that an AI like Claude or Llama2 was trained on (many open-source LLMs have indeed ingested Stack Exchange data). Later, if someone asks the AI, “What are some reliable cloud hosts for X use case?”, the model might recall the solutions mentioned frequently – including that company from Stack Overflow. Similarly, a highly upvoted Reddit comment describing the pros and cons of a new smartphone might influence an AI’s answer to “Is [Smartphone XYZ] any good?”. Authentic Participation: The key word here is authentic . Online communities can spot overt marketing a mile away, and heavy-handed self-promotion can backfire (and may even violate platform rules). The goal is to contribute value : answer user questions, provide insights, share knowledge freely. It’s acceptable to disclose your affiliation (in many forums it’s required), as long as you’re genuinely helping and not just plugging your brand. In fact, being transparent about your company role while offering useful info can build goodwill . Other users may start to mention your brand positively even when you’re not in the conversation. For instance, representatives from tech companies often participate in subreddit AMAs (Ask Me Anything) or help troubleshoot issues on forums – this humanizes the brand and generates content that could later serve as training data. As one SEO expert noted, “Engaging authentically on behalf of a brand (with affiliation revealed) is often welcomed to clarify user questions, and it will likely benefit your generative AI optimization journey too.” ( ^[15] ). In short, if people are talking about your brand (and in the right way) on forums, then AI will eventually talk about your brand as well . Social Media Signals and AI Models: Social networks like Twitter (now X), LinkedIn, YouTube, and TikTok represent another facet of off-page presence. Traditionally, Google has maintained that social signals (likes, shares) aren’t direct ranking factors. But for LLMs, the relationship is different: the content of social media can become part of the AI’s knowledge. For example, OpenAI’s GPT models up to GPT-4 were trained on a portion of the public web which likely included some Twitter content (especially before certain dates). Meanwhile, Elon Musk’s new AI venture xAI explicitly uses public Twitter/X data to train its chatbot Grok ( ^[16] ) ( ^[17] ). Grok is designed for real-time, up-to-date responses and “integrates real-time learning from X (Twitter) to provide context-aware answers.” ( ^[18] ). This means that what’s trending or frequently discussed on X can influence Grok’s output. If your brand is a frequent subject of tweets – say tech influencers often mention your product – Grok might “know” a lot about it. Conversely, a brand with no Twitter presence or discussion might be invisible to Grok except for whatever it scrapes from the web. It’s not just Grok. Other models and search AIs are likely to incorporate social content in various ways: Google has hinted at incorporating more “voices” from forums and social in its results (hence the SGE “perspectives” section). Future Google AI could use public social content to add nuance or opinions to answers. Bing integrates some Live Search and could theoretically pull in social posts if relevant (especially for real-time queries, e.g., Bing’s chatbot might show a live tweet for a breaking news question). Localized AI models (like Baidu’s ERNIE in China) might ingest data from their local social platforms (Weibo, Zhihu, etc.). We’ll discuss international specifics later, but the concept holds: in any market, the prevalent social/community platforms influence that locale’s AI knowledge. Being Present and Valuable: So how can a brand leverage social and community presence for AI benefits? Establish expertise on Q&A platforms: If you are in B2B, for instance, answer questions on Quora with depth and objectivity. If Quora content is used by an AI (some models likely trained on Quora answers), your explanations might teach the AI about concepts and connect them to your brand. It’s not about promoting, but about being part of the informational canon. Foster discussion in communities: Engage with communities like Reddit in your niche. Some companies create official accounts or even brand-specific subreddits to host discussions. Even broader participation helps. Imagine a popular Reddit thread “What’s the best mattress for back pain?” If a lot of users happen to mention a particular D2C mattress brand (because the brand seeded a few trial units to Redditors or simply has a good product that people naturally recommend), that thread’s content might later shape an AI’s answer to “Recommend a mattress for back pain.” Leverage YouTube and visual platforms: YouTube comments and transcripts are often indexed by search and possibly consumed by AI training. If your brand has a strong presence on YouTube (through educational videos, for example) and those are discussed or referenced widely, it adds to the off-page signal. Additionally, AI models are evolving to be multimodal – they might parse video transcripts or audio in the future. Already, voice assistants and AI can parse YouTube transcripts for info . A tutorial video where an influencer reviews your product favorably could indirectly influence AI answers to related questions (“How to do X” – AI might recall the method described in the video). Be mindful of sentiment: Community and social content often include subjective opinions – which AI might pick up. If many users complain about a specific flaw of your product on forums, an AI could surface that as a “common con” when asked about your product (because it has effectively done a tally of what people say). Inversely, if a particular feature is consistently praised in user discussions, the AI is likely to mention that pro. This connects to the next section (reviews), but the principle is: widespread social sentiment can become AI “knowledge.” Brands should therefore monitor and engage in these spaces not just to push positive messages, but to address issues. Demonstrating customer support publicly (responding to a complaint on Twitter, for example) can even turn a negative into a positive knowledge point (“Brand X is known for responsive support”). Reddit and others as cited sources: Interestingly, some AI search engines cite Reddit or Stack Exchange directly in answers. Perplexity.ai , an AI search assistant, often provides answers with footnotes from sources, and it’s not uncommon to see a Reddit URL in those footnotes for certain queries (especially technical or niche questions). Google SGE, in its early 2023 demo, showed an example of an AI snapshot answer that included info likely drawn from forums (Google also launched a “Perspectives” filter to highlight forum/social media answers alongside the AI summary). The implication for marketers is: the content you or your users create on community platforms can get directly quoted or surfaced by AI . Imagine hosting a highly informative thread on your own community forum – Google’s AI might actually pull a sentence from a user’s post there into the answer box (with a citation). If that happens, it’s as good as an organic ranking, even if it’s not on your main site. Thus, investing in community building (like forums, user groups, Discord servers) not only helps traditional community engagement but could become part of the AI answer ecosystem. International Note – Community Presence Beyond English: Different markets have their own dominant platforms. For example, in China , people turn to Zhihu (a Quora-like Q&A site) or Weibo for discussions, and Baidu’s search AI will likely tap into Baidu Zhidao (Q&A) and Baidu Tieba (forums). In Russia , VK and local forums matter. Ensure that your international marketing teams are engaging where local users ask questions. If your brand has a global footprint, being present in those native-language discussions is crucial – not only will it reach human audiences, it will be reflected in any region-specific AI models or search engines. Many non-English LLMs (and multi-lingual ones) have training data drawn from local Wikipedia editions and forums, so raising your brand profile in those languages (through PR and community talk) feeds the model similar to English content. To summarize, social and community presence is the new “word-of-mouth” in the AI training corpus . What humans say in these spaces becomes what the machine repeats. By actively participating and fostering positive, informative conversations about your brand on these platforms, you increase the chances that when an AI is answering a related question, your brand’s perspective or name will surface . It’s a long-game, largely indirect, but it aligns with an authentic marketing approach – helping people in the channels they already frequent, which in turn helps the AI pick up your brand’s trail.

11.3 Reviews and User-Generated Content: The New Influencers of AI Recommendations

When consumers are looking for products or services, reviews and user-generated content (UGC) play a pivotal role. In the AI era, this dynamic remains – but now the LLMs themselves act as intermediaries , synthesizing and conveying the collective voice of users. If you want your brand to be recommended or favored by AI in contexts like “What’s the best [product] for [need]?”, earning positive reviews and fostering UGC is critical. Models trained on e-commerce data, forums, and Q&A sites will pick up on these signals (like “best rated” , common pros and cons, etc.) when formulating answers. UGC as Training Data: Think about what an LLM knows regarding a product category. A large portion of that knowledge might come from scraping sites like Amazon (product descriptions and possibly some reviews), Trustpilot, TripAdvisor, CNET, niche review blogs, and more. Even if not intentionally included, models might ingest review content that’s prevalent on the open web. For example, a generative AI might have learned that “Restaurant ABC has a 4.7/5 rating and people often praise the ambiance but mention slow service” simply because those points appeared across dozens of Google Reviews or Yelp snippets that made their way into some web crawl. Now, when a user asks “Is Restaurant ABC good for a date night?”, an AI could respond, “It’s highly rated (4.7) and known for great ambiance, though some reviews mention the service can be slow.” In doing so, it has effectively echoed user-generated sentiment . We already see glimpses of this in how AI chatbots answer product queries: ChatGPT with browsing or plugins : OpenAI has experimented with a browsing mode and plugins that can pull real-time info. If you ask, for instance, “What are the pros and cons of the XYZ camera?”, a browsing-enabled AI might fetch a page like a Best Buy customer reviews summary or an expert review, then tell you “Pros: excellent image quality, durable build; Cons: expensive, heavy – as noted by many reviewers on BestBuy.com.” Even without browsing, base GPT-4 may have seen enough commentary to enumerate common pros/cons (with some risk of hallucination if not retrieved live). Bard (Google’s AI) : It can draw on Google’s Knowledge Graph which includes aggregated review info. Bard or SGE might actually say “This vacuum is rated 4.5 stars, with users liking its suction power but disliking the noise.” That kind of response directly reflects UGC. Bing Chat : Similarly, when asked about “best laptops under $1000”, it might cite sources like LaptopMag or Reddit threads, often incorporating consensus opinions (“most users find the battery life of Model X to be excellent”). One telling example came from the earlier scenario with HubSpot. When ChatGPT-4 was prompted to give pros and cons of HubSpot for small business marketing, it did so and the cons it listed correlated with points from a NerdWallet review page ( ^[19] ). The AI’s citations even pointed to NerdWallet (an affiliate site known for product comparisons) as a source. This shows that AI will mine third-party reviews and comparisons to answer specific brand questions , effectively acting like an ultra-fast review aggregator. If your product has a recurring negative mentioned across many reviews, don’t be surprised if an AI brings it up. Conversely, if you have genuinely rave reviews highlighting a strength, the AI will likely emphasize that point. Encourage Positive Reviews (Genuinely): No, this doesn’t mean spam the system with fake 5-stars – AI may eventually catch patterns of inauthenticity and it certainly violates ethics. Instead: Provide great products and service so that positive reviews happen organically. (It always starts here!) Invite feedback from satisfied customers on platforms that matter (Google, Amazon, industry-specific sites). Often, a steady stream of recent positive reviews not only boosts SEO and conversion but also seeds the AI’s “memory” with favorable content. Address negative feedback openly and promptly. If you can turn a 2-star customer into a 4-star by solving their issue, that follow-up might become part of the narrative that an AI learns: “Customers initially had issues with X, but the company was quick to resolve them.” This could mitigate an otherwise damaging generalization. Structured Data for Reviews: There’s a technical side to this as well. By using schema markup (Review and AggregateRating schema) on your site’s testimonials or product pages, you feed structured information to search engines and potentially to AI. Google’s SGE has shown product summary boxes that include star ratings and price ranges. These are pulled from structured data and Google’s own shopping data. If AI has access to that, it will use it. Also, companies like OpenAI and others might use schema-structured data as part of training or retrieval for factual consistency. The bottom line: make sure your high-level review info (average rating, number of reviews) is easily machine-readable. It helps traditional SEO (rich snippets) and gives AI factual fodder. Models Trained on E-commerce Data: One emerging area is AI-powered shopping assistants . OpenAI launched a ChatGPT Shopping feature in 2025 that integrates with Shopify and uses real product data to make recommendations ( ^[20] ) ( ^[21] ). Early indications show it leverages reviews data as a primary source of truth when deciding what to recommend. Adobe’s e-commerce reports found that traffic to retail sites from generative AI sources rose 1200% from mid-2024 to early 2025 as these tools rolled out ( ^[21] ). That’s an astronomical jump (albeit from a small base) and signals that AI shopping helpers are becoming a real referral channel . To “win” in those interactions, a brand needs to look attractive in the data the AI sees – which includes review sentiment, ratings, and product specs. A 2025 guide from Reviews.io (a review platform) emphasizes how high-quality review content can boost AI-driven product visibility : “Reviews offer the kind of high‑quality structured data that generative AI engines love.” ( ^[22] ). The guide suggests that it’s not just having lots of reviews, but having the right kind of detail in them . For instance: Verified reviews add trust (AI can distinguish verified purchases, which carry more weight). Reviews that mention specific attributes or use-cases (“this stroller is great for traveling because it folds compactly”) provide context-rich content that AI can latch onto ( ^[23] ). Questions and answers (Q&A content) on product pages – often UGC where customers ask something and the brand or community answers – give clear, concise info that AI can directly use. One interesting tactic is excerpting and summarizing reviews in a way that machines can easily parse. Some brands use widgets that compile common pros and cons from multiple reviews (like a summary: “ Pros: long battery life, lightweight; Cons: limited color options”). This effectively does the AI’s job of aggregation, and if it’s visible on your site (marked up properly), an AI might pick that up. In fact, Reviews.io offers an “Expert Answers” Q&A widget and review tagging for themes ( ^[24] ) ( ^[25] ) – all aimed at structuring UGC so AI can digest it. The mantra is: treat your reviews as data for AI . High volume of authentic, detailed reviews = robust data. Structure and label that data = easier for AI to learn from it or retrieve it. Influencer Content as UGC: Don’t overlook user content beyond text reviews. YouTube reviews, Instagram testimonials, TikTok unboxings – these are UGC too. As AI becomes multimodal, it might interpret spoken/written words in videos or images. Already, Bing’s AI can summarize YouTube videos upon request; tomorrow’s AI might integrate that knowledge proactively. So encouraging customers or influencers to create content around your product can indirectly feed the AI narrative. If many tech influencers say “Battery life of this laptop lasts ~8 hours” in their videos, an AI that’s connected to a transcription service might state that fact when asked. This is speculative but within the realm of possibility given how fast AI capabilities are evolving with multimedia. UGC on Your Own Properties: A quick note on hosting user-generated content on your site (like comments, forums, Q&A). This can be double-edged. On one hand, it creates on-page content that could rank (traditional SEO) and could be picked by AI as part of your page content if it’s retrieving from your site. On the other, if not moderated, it can include incorrect info or spam which you wouldn’t want propagated by an AI. So, moderate carefully. Foster a positive community whose content can serve as a knowledge source. Case in Point – The “Best Product” Queries: Everyone wants to be the answer when someone asks an AI “What’s the best [your category product]?”. AI tends to answer such questions with a list of a few options (often 3-5) with brief rationales. What determines those picks? Based on observed behavior in 2024-2025: Brands frequently recommended on affiliate “best” lists (the AI might have seen 10 different “Top 10 X” articles and noticed certain names keep appearing at the top). Products with very high average ratings (if the AI has access to aggregate ratings, it may favor those above a threshold). Products with distinguishing features that stand out in reviews (the AI loves to justify its choice with something – “X is best for budget-conscious buyers, Y is best overall for quality,” etc. It learns those associations from the way people talk about the products). So, if you are not at least in the conversation on those fronts, you’ll be absent from the AI’s answer. This might require not just one tactic but a synergy: get into the review roundups (PR), ensure your product truly satisfies customers (product dev), encourage them to leave reviews highlighting why they chose you (marketing/CRM), and even subtly highlight those strengths in your own content so that they get quoted. To illustrate, imagine Company A and Company B both make a smart home device. Company A has a 4.8-star rating with hundreds of reviews mentioning “easy to set up” and “great customer support.” Company B has a 4.2-star rating with some reviews praising advanced features but many mentioning “glitchy software.” If a user asks the AI, “Which smart home hub should I buy?”, the AI might answer: “You have a few options. The XYZ Hub (Company A) is highly rated for its ease of setup and reliable support, which makes it a great choice for most people ( ^[24] ). Alternatively, ABC Hub (Company B) offers more advanced features, but some users report software issues.” Here, clearly, Company A wins the recommendation. Why? Because the user consensus in UGC favored it, and the AI mirrored that consensus. Managing and Monitoring Reviews for AI Impact: It’s now important to monitor your reviews not just for human reputation but for AI portrayal . Are there misconceptions propagating in reviews? (e.g., multiple people wrongly assuming your product lacks a feature and complaining – the AI won’t know they’re wrong; it will assume your product indeed lacks it because “many users say so.”) If so, you might need to address it in an FAQ or response so that corrected information exists for the AI to find. Also, keep an eye on how AI currently describes your product. You can literally ask ChatGPT or Bing, “What do people like about [Product] and what do they dislike?” See if the answer aligns with your understanding. If an AI says something like “Many people say it’s overpriced,” you know that’s out there in the collective feedback and should be a concern to tackle (either via product pricing strategy or via communicating value better). Lastly, UGC extends to broader discussions about your brand outside of dedicated review channels . This overlaps with social/community presence: a Reddit thread “Anyone try [YourApp]? What’s your experience?” is effectively a user review discussion. These informal discussions can influence AI too. Encouraging happy customers to share their experiences in forums or communities (when appropriate) can create positive “grassroots” UGC that shapes perception. In summary, user-generated content – especially reviews – are the lifeblood of an AI’s recommendations . They represent the voice of the customer at scale, and AI listens to that voice attentively. By ensuring your off-page presence includes a strong constellation of positive, authentic reviews and by structuring that content for machine consumption, you greatly enhance your chance of being the brand that AI suggests when it’s playing the role of salesperson or advisor.

11.4 Collaboration, Knowledge Sharing, and Building a Web of Influence

Another powerful way to boost off-page authority in the AI era is through collaboration and knowledge sharing. This means contributing value to the broader industry ecosystem – via research, data, open-source projects, or educational content – such that your brand becomes ingrained in the collective knowledge that AI draws upon. When your insights or resources get referenced by other websites, analysts, or even Wikipedia, you effectively feed the AI new information with your brand’s name attached. Over time, these contributions make your brand an “entity” that AI models recognize and respect. Contribute to Industry Research & Whitepapers: Many companies have started to publish original research or partner with academic institutions to conduct studies. If your organization produces a well-regarded annual report (for example, “State of Remote Work 2025” or “Cybersecurity Benchmarks Report”), this can get widely cited. Not only does it earn backlinks (traditional SEO win), but it also means any AI trained on 2025 web content will very likely ingest parts of that report and the fact that “according to [Your Company’s] research, XYZ.” If that stat is useful, AI might even quote it in answers. In fact, LLMs often preface facts with “according to [Source]…” if their training included that phrasing. Imagine a user asks, “What percentage of companies plan to increase AI spending next year?” If your company’s whitepaper found “45% plan to increase AI spend,” an AI might reply, “According to a 2025 report by [Your Company], 45% of companies plan to increase AI spending next year.” – This directly inserts your brand into the AI-generated answer (and establishes you as an authority on that topic). For this to happen, your research needs to be openly accessible (at least key findings) and preferably covered by secondary sources (news sites writing about your report).

A mention in high-authority contexts – e.g., your stat gets into a Wikipedia article or a Forbes piece – further solidifies it. There’s a virtuous cycle: you share knowledge → others cite it → AI sees multiple citations and trusts/learns it. “If your data or insights get referenced by other websites or Wikipedia, they are more likely to become part of the corpus AI learns from,” as our outline noted. A single Wikipedia mention could mean an AI will “know” your brand or findings and might cite them when relevant in conversations. Open Data and Open-Source Projects: In tech fields especially, contributing to open-source or open data initiatives can amplify your off-page presence. For example, if your company open-sources a useful software library on GitHub, developers worldwide might use and discuss it. GitHub discussions and stars are one indicator (some LLMs have knowledge of popular GitHub repos up to a point). Also, documentation might be indexed. There’s evidence that Meta’s Llama model, for instance, internalized a lot of knowledge from coding sites and GitHub. If your brand name is tied to a popular open-source tool (think of “Facebook’s React library” or “Google’s TensorFlow” – those brands get huge authority by association in AI’s mind), that’s a massive boost in AI-era clout. Likewise, releasing open datasets or participating in community data challenges (Kaggle competitions, for example) can get your brand name circulating among practitioners and in published solutions. If researchers use your dataset and credit your company in papers or blogs, that’s creating intellectual backlinks. LLMs that read arXiv papers or blogs may pick up those references. Even something like being listed as a data provider in a prominent dataset index could raise your profile.

Knowledge Sharing via Wikis and Forums: Many companies also share knowledge via channels like Medium posts, developer forums, or knowledge hubs. If done in a way that others link to it as a reference, it’s beneficial. One notable angle is Wikipedia editing. While directly writing your own Wikipedia page is discouraged (conflict of interest), you can legitimately contribute to Wikipedia on topics you know, citing reliable sources (not your own marketing material, but perhaps your research). If your work is truly notable, someone else might add it. Having your brand or product mentioned in relevant Wikipedia articles (not just the page about your brand, but in the context of some technology or concept) can be huge. For instance, the “ChatGPT” Wikipedia page might list companies that have partnered with OpenAI – being on that list means every clone of Wikipedia (and there are many in datasets) includes your name. Entity Recognition and Knowledge Graphs: Collaboration and being cited feeds not just the raw text AI, but also the structured knowledge systems like Google’s Knowledge Graph or Microsoft’s LinkedIn-based graph. If you contribute to standards bodies, have executives on association boards, or partake in notable collaborations, you become part of those networks of information. Google’s Knowledge Graph, for example, might give your brand an “entity” with various connections (CEO name, founded date, industry, key product). Bard and SGE likely consult that for factual consistency. Ensuring that public knowledge about those facts is accurate (via Wikipedia/Wikidata or authoritative sites) is important. One can add or edit entries in Wikidata (which is a structured database many AIs use for grounding). Ensure your company’s Wikidata entry is up to date with proper references. It’s a behind-the-scenes task that could pay off when AI needs to spit out your headquarter location or tagline correctly.

Collaborate with Influencers and Thought Leaders: Another form of collaboration is co-creating content with respected voices. For example, co-author a paper with a university lab, or have a well-known expert contribute a chapter to your eBook (like this!). When those experts mention the work, it adds credibility and visibility. AI models might not know every niche influencer, but they certainly know some (especially if those people have Wikipedia pages or are cited frequently). If, say, a famous AI professor tweets about your research and blogs it – that might indirectly propagate into the AI’s sphere. The AI might not know your brand initially, but it knows the professor is trustworthy (from their published work), and now that professor’s blog discussing your research becomes a source the AI trusts. Community Knowledge Sharing: On a smaller scale, sponsoring or actively participating in community knowledge bases (like answering on Stack Exchange, writing in Medium Publications, speaking in webinars or podcasts that get transcribed) spreads your brand’s expertise. Many of these content forms (Q&A answers, blogs, transcripts) become web text that LLMs train on or search engines index. For instance, if your CTO answers many questions on Stack Overflow about a certain programming language, an AI might one day answer a user query with, “According to [Your CTO’s Name] on Stack Overflow, the best way to optimize in that scenario is to do X.” It sounds far-fetched, but we’ve already seen ChatGPT cite individual’s answers from forums in some plugin-enabled modes. Educational Resources and MOOCs: If relevant, create free educational content – like a mini-course, certification, or tutorial series. If widely used, this can become part of the canon.

E.g., HubSpot’s free marketing academy got their terminology and frameworks into widespread use, some of which even AI might reference (e.g., “According to HubSpot’s Flywheel model…” etc.). Offering value without paywall makes it more likely external sites will reference it (and thus AI too). In short, sharing knowledge freely and collaborating beyond your own walls builds an external web of references to your brand. It positions your company as a thought leader or at least a contributor to the advancements in your field. AI models, which are essentially giant prediction machines, lean on these bread crumbs of facts and references to form their answers. By increasing the density and quality of those bread crumbs with your name on them, you create pathways for the AI to mention or credit you. One more angle: public data contributions. Some organizations have started putting datasets or insights on public data portals or even contributing to government or NGO research. For example, a fintech startup might provide anonymized trend data to a central bank’s annual report. If cited, that could later be reflected in any AI summarizing the report. Or a healthtech company might share statistics with the World Health Organization for an advisory – later an AI might say “WHO data indicates X” (and the fine print source was originally that company’s contribution). Lastly, monitoring this aspect is tricky – how do you know if your collaborations are noticed by AI? One way is to see if your brand or key personnel have made it into knowledge base entries (like Google’s Knowledge Panel or Wikipedia). Also, simply ask AI models: “What do you know about [Company Name]’s research on [Topic]?” or “Does [Company] have any open-source projects?” If the AI can answer with specifics that you’ve indeed put out there, it’s working.

If it says “I’m not aware” or answers incorrectly, that shows a gap in the AI’s training recognition of your contributions – maybe your stuff didn’t circulate widely enough or is too recent to have been ingested. That’s feedback to either promote it more or ensure it’s indexed. Global and Non-English Collaboration: Ensure that your knowledge sharing isn’t confined to English if you operate in other languages. For instance, publishing a Spanish version of your whitepaper can lead to Spanish language sites citing it, thus Spanish LLMs or multilingual models getting it. Some countries have local encyclopedias or knowledge bases (e.g., Baidu Baike, the Chinese Wikipedia equivalent). If your brand is global, consider working with local partners or agencies to establish a presence in those. Being part of an international research effort or standard (like W3C, ISO, etc.) can also put you on the map globally in AI’s eyes, as those are highly respected sources. In essence, think of collaboration and knowledge sharing as seeding the information landscape. It’s planting seeds of data and insight that others water by referencing, and AI harvests when generating answers. Unlike SEO where sometimes content stays siloed on your site, here you want your ideas to spread and be rehosted or mentioned elsewhere (with credit ideally). It’s a slightly altruistic approach – give more to get more presence – but it can yield high authority returns that no amount of on-page keyword tweaking could achieve.

11.5 Monitoring Brand Mentions in AI Outputs

After investing in off-page SEO and brand authority efforts, it’s crucial to measure their impact in the AI landscape. Traditional SEO has an arsenal of tools for tracking rankings, backlinks, and mentions on the web. Now, analogous tools (and methods) are emerging to track brand mentions in AI-generated content – essentially keeping an eye on how and when your brand appears in responses from chatbots and generative search results. Monitoring this helps you understand your current visibility and reputation in AI outputs, and it provides feedback for further optimization. Manual Testing with AI Chatbots: A simple starting point is to ask the AI directly about your brand. As recommended by SEO experts, use generative AI tools themselves to learn what they know (or don’t know) about your brand ( ^[26] ). For example: “Tell me about [Your Company].” – Does the AI give a correct summary? Does it cite sources? Are those sources your site or Wikipedia or something else? “What are the best companies/products in [your industry]?” – Does your name appear? If not, who is dominating and why? “What are the pros and cons of [Your Product]?” – What negatives does it list, and what sources is it drawing from (reviews, competitor comparisons, etc.)? “Is [Your Brand] reputable for [service]?” – This can surface any trust or quality issues the AI might have picked up (maybe from reviews or news). Alli Berry, in the Search Engine Journal article we examined, did this for HubSpot ( ^[6] ) ( ^[27] ). She found that ChatGPT’s answer cited HubSpot’s own legal pages and KB articles, and in a “best brands” query it cited Wikipedia and NerdWallet for context ( ^[3] ). By drilling down with follow-up questions (pros and cons of HubSpot), she could identify exactly which review sites or sources were feeding the AI information ( ^[19] ). This is invaluable. It’s like reverse-engineering the AI’s brain to see which content pieces influenced its knowledge. If you find the AI referencing certain articles or sites frequently in answers about your space, those are sources you either want to be present on or improve your presence on. Perhaps a specific review blog holds a lot of sway (as NerdWallet did in that case); you might consider engaging that site for an updated review or partnership so that future AI answers have better info. Alli’s approach and advice: If the AI cites an obscure or less credible site for info on you, that’s an opportunity to provide better, more authoritative content on the topic, or to get coverage from a more authoritative site ( ^[28] ). She noted that some citations were not well-known, so PR could aim to get more authoritative coverage, which hopefully the AI would incorporate next time ( ^[29] ). If the AI lists cons that have validity, feed that back into product improvement, and build relationships with the sources that mention those cons so they can update their content when you address the issues ( ^[30] ). Essentially, treat AI as another channel where your brand reputation is playing out, and manage it accordingly. Tools for AI Brand Monitoring: As the importance of AI visibility grows, several SEO and analytics companies have rolled out features to track this. For example: Ahrefs’ Brand Radar – A feature launched in 2025, which lets you track brand mentions in ChatGPT and Perplexity , with Google’s Gemini promised to be included as well ( ^[31] ). It provides an index of questions where your brand appears in AI answers and how you show up. This helps quantify, for instance, “Our brand was cited in 25 different ChatGPT answers last month” – something that was impossible to know before. It’s updated regularly, showing trends over time. Authoritas / SGE Monitor – Some tools focus on Google’s Search Generative Experience. Authoritas claims to have an AI Overview rank tracker that not only detects when an AI summary appears for a keyword, but whether your site was cited in it and at which position ( ^[32] ). This is akin to rank tracking but for AI inclusion. If you optimize a page and suddenly it starts getting cited in SGE, these tools would catch that. Keyword.com’s AI Rank Tracker – A platform that touts tracking across multiple generative AI platforms (ChatGPT, Google SGE, Perplexity, Claude, even DeepSeek and Mistral – basically any notable LLM chat/search) ( ^[33] ) ( ^[34] ). It logs which queries produced an answer that mentions your brand and in what context. They often include features like prompt tracking (seeing what questions trigger your brand) and citation analysis (which of your pages or content are getting cited) ( ^[35] ) ( ^[36] ). There’s also an aspect of sentiment analysis – some tools try to gauge if those mentions are positive, neutral, or negative. Specialized Monitoring Tools – Besides the big SEO suites, some startups and SaaS tools have emerged purely for AI monitoring. For example, PromptMonitor.io or the ones mentioned in search results (like Irene Chan’s list of 9 tools ( ^[37] )) and others on Reddit. These might allow you to input your brand name and then periodically query various AI models via API to see how you’re mentioned. It’s like doing the manual Q&A method but automated and at scale. Tracking AI Referral Traffic: Another angle is to look at your web analytics for traffic coming from AI tools. For instance, Bing Chat and ChatGPT (with browsing) may actually send traffic if users click the citations. In GA4 or server logs, you might see referrals from domains like bing.com with unusual parameters, or from chat.openai.com (if someone using the browsing tool clicked your link). One SEO shared that they saw referral traffic from perplexity.ai and chat.openai.com after allowing those bots to crawl ( ^[38] ). Setting up segments or alerts for such referrals can indicate, indirectly, that you were mentioned and someone clicked through. Granted, if AI answers get so good that users don’t click, you might not get traffic even if you were mentioned – but tracking a bump or drop in these new referral sources is useful to gauge AI impact. GA4 can be configured with custom events to track, say, “Visits from AI summaries” if you tag those referral patterns ( ^[39] ) ( ^[40] ). This is bleeding-edge analytics; not all companies are doing it yet. Monitoring Brand Sentiment in AI: It’s not just whether you’re mentioned, but how . Some tools and studies look at sentiment analysis of AI mentions ( ^[41] ) ( ^[42] ). For example, are you being recommended enthusiastically (“highly regarded”) or cautiously (“some customers report issues with…”)? If an AI consistently frames your brand negatively, that’s a red flag that the source information it has is skewed negatively (or possibly the AI has hallucinated something harmful). Either way, you’d want to investigate and counteract: maybe bolster positive content, address the root cause of negatives, or provide clarifications. There is even talk of a concept called “narrative anchoring” , where the first exposure an AI got about your brand (like an old news article or a big controversy) can anchor its narrative unless new information supplants it ( ^[43] ). Monitoring tools that highlight the descriptors or common context around your brand in AI answers can help catch that. For instance, if the AI always mentions a 2023 data breach whenever your brand comes up, you need to flood the ecosystem with more current info and positive developments to dilute that association. Competitive Insights: While monitoring your brand, don’t forget to also monitor key competitors in the same way. Many AI tracking tools let you watch multiple brands. If you notice competitor X is constantly mentioned by ChatGPT in answers where you are not, analyze why. It could be their stronger off-page presence or a successful PR campaign. This can inform your strategy: you might need to mirror their tactics (e.g., get into the same review articles or get a Wikipedia page if they have one, etc.). Conversely, you might discover gaps – maybe the AI gives outdated info about a competitor (like a product feature they no longer have). That’s an opportunity: if even the AI is behind, maybe that competitor hasn’t been active in sharing updates, and you can take advantage by being more present with updated content (making you the fresher source). Staying Updated with AI Platform Changes: Keep an eye on updates from OpenAI, Google, Microsoft, etc., about how their AI systems incorporate content. For example, OpenAI’s policies around its web crawler GPTBot evolved – initially many sites blocked it (concerned about their content being used without credit), but if you choose to allow it, you’re essentially agreeing to let ChatGPT train on your site’s content. As of 2024, about one-third of top websites blocked GPTBot ( ^[44] ) ( ^[38] ), including some big brands. If your competitors block it and you don’t, your content might feature more prominently in future GPT models (because their content wasn’t included). It’s a strategic decision: visibility vs. intellectual property concerns . Either way, being aware of these developments is part of monitoring the landscape. Direct Feedback from AI Companies: While not common yet, we might see tools from the AI providers themselves. OpenAI has experimented with a “Browse with Bing” that cites sources . Google’s SGE is citing and might eventually show site owners data in Search Console about AI appearances (nothing official as of early 2025, but logically they might). Bing Webmaster Tools could provide something on how often your site was used in Bing’s chat answers. So, keep an ear out for such features – they would greatly ease monitoring if/when they arrive. In practice, a monitoring routine could be: Every month, use ChatGPT, Bard, and Bing to run through a list of key brand queries (the ones we listed earlier). Document the responses and sources. This is a manual “audit” of your AI presence. Use an automated tool (if budget allows) like Ahrefs or Keyword.com to continuously watch a broader set of queries and send alerts when something changes (e.g., your brand appears for a new query or disappears from one). Track referral traffic and set up alerts for spikes or drops from domains associated with AI (OpenAI, Bing, Perplexity, etc.). Regularly prompt the AI in a neutral way: “What is [Brand] known for?” and see if any misinformation creeps in. If yes, that’s a crisis prevention flag – you might need to do damage control in the source content or via PR. Using AI to Audit Content: Ironically, you can also use AI itself to help with monitoring. For instance, you could use a tool like GPT-4 to analyze a large set of AI responses about multiple brands and have it summarize comparative mentions. Or use Python with AI APIs to simulate hundreds of user questions and parse the answers for mentions of your brand. This is more technical, but some advanced SEO teams are doing such programmatic analysis. Finally, treat the insights from monitoring as actionable data. If you find, for example, that: You’re not being mentioned where you should be – revisit Sections 11.1–11.4 to amplify off-page efforts in those weak spots. You are being mentioned but incorrectly – consider an outreach to correct the sources (e.g., a wiki edit if factual, or contacting a blogger who had outdated info). A competitor gets a lot of love from AI – analyze their off-page footprint and emulate or outdo it. Users via AI are asking something related to your domain that no one (neither you nor competition) has covered well – perhaps create that content and propagate it so you become the go-to reference (first-mover advantage in AI answers). The era of “AI SEO” or GEO (Generative Engine Optimization) is still new and evolving. Monitoring is how you keep your finger on the pulse and adapt quickly. As one expert said, “We will have to be more reliant on ourselves to reverse-engineer what we’re seeing in the data and run our own experiments” ( ^[45] ). By actively checking how AI portrays your brand, you essentially close the loop on your off-page strategy – ensuring all the brand authority you’re building externally is indeed translating into the AI-driven search results of the future.

References

^[1] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/posts/pete-buckley_llms-love-brands-but-ignore-brand-advertising-activity-7338115418407989248-2KuB

^[2] Sparktoro.Com Article - Sparktoro.Com URL: https://sparktoro.com/blog/how-can-my-brand-appear-in-answers-from-chatgpt-perplexity-gemini-and-other-ai-llm-tools

^[3] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[4] Sparktoro.Com Article - Sparktoro.Com URL: https://sparktoro.com/blog/how-can-my-brand-appear-in-answers-from-chatgpt-perplexity-gemini-and-other-ai-llm-tools

^[5] Sparktoro.Com Article - Sparktoro.Com URL: https://sparktoro.com/blog/how-can-my-brand-appear-in-answers-from-chatgpt-perplexity-gemini-and-other-ai-llm-tools

^[6] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[7] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[8] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[9] Aclanthology.Org Article - Aclanthology.Org URL: https://aclanthology.org/2024.emnlp-main.707.pdf

^[10] Aclanthology.Org Article - Aclanthology.Org URL: https://aclanthology.org/2024.emnlp-main.707.pdf

^[11] Aclanthology.Org Article - Aclanthology.Org URL: https://aclanthology.org/2024.emnlp-main.707.pdf

^[12] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/posts/pete-buckley_llms-love-brands-but-ignore-brand-advertising-activity-7338115418407989248-2KuB

^[13] www.theclueless.company - Theclueless.Company URL: https://www.theclueless.company/llm-seo

^[14] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[15] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[16] Medium.Com Article - Medium.Com URL: https://medium.com /@serverwalainfra/grok-ai-and-real-time-learning-how-it-leverages-x-for-up-to-date-responses-01d7148fc041

^[17] Medium.Com Article - Medium.Com URL: https://medium.com /@serverwalainfra/grok-ai-and-real-time-learning-how-it-leverages-x-for-up-to-date-responses-01d7148fc041

^[18] Medium.Com Article - Medium.Com URL: https://medium.com /@serverwalainfra/grok-ai-and-real-time-learning-how-it-leverages-x-for-up-to-date-responses-01d7148fc041

^[19] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[20] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[21] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[22] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[23] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[24] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[25] Reviews.Io Article - Reviews.Io URL: https://blog.reviews.io/post/chatgpt-shopping-strategy-improve-product-rankings-with-review-data

^[26] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[27] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[28] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[29] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[30] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

^[31] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[32] www.theclueless.company - Theclueless.Company URL: https://www.theclueless.company/llm-seo

^[33] Keyword.Com Article - Keyword.Com URL: https://keyword.com/ai-search-visibility

^[34] Keyword.Com Article - Keyword.Com URL: https://keyword.com/ai-search-visibility

^[35] Keyword.Com Article - Keyword.Com URL: https://keyword.com/ai-search-visibility

^[36] Keyword.Com Article - Keyword.Com URL: https://keyword.com/ai-search-visibility

^[37] Irenechan.Co Article - Irenechan.Co URL: https://irenechan.co/monitor-chatgpt-brand-mentions-platforms

^[38] www.reddit.com - Reddit.Com URL: https://www.reddit.com/r/TechSEO/comments/1ladbhr/ai_bots_gptbot_perplexity_etc_block_all_or_allow

^[39] www.theclueless.company - Theclueless.Company URL: https://www.theclueless.company/llm-seo

^[40] www.theclueless.company - Theclueless.Company URL: https://www.theclueless.company/llm-seo

^[41] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/posts/pete-buckley_llms-love-brands-but-ignore-brand-advertising-activity-7338115418407989248-2KuB

^[42] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/posts/pete-buckley_llms-love-brands-but-ignore-brand-advertising-activity-7338115418407989248-2KuB

^[43] www.withdaydream.com - Withdaydream.Com URL: https://www.withdaydream.com/library/protect-your-brand-in-the-age-of-ai-search

^[44] www.stanventures.com - Stanventures.Com URL: https://www.stanventures.com/news/major-websites-block-openais-gptbot-amid-privacy-concerns-521

^[45] www.searchenginejournal.com - Searchenginejournal.Com URL: https://www.searchenginejournal.com/how-to-get-brand-mentions-in-generative-ai/539570

12. Prompt Optimization and Ethical Influence on AI Outputs

12.1 Understanding User Prompts in the Age of AI Search

The way people search for information is shifting from typing keywords into a browser to asking full-fledged questions in conversational AI tools. A few years ago, a typical query might have been “best running shoes 2022” —terse and keyword-based. Today, users are more likely to ask an AI assistant something like: “Which running shoes are best for long-distance runners with flat feet under $150?” This represents a fundamental change in search behavior. People treat AI assistants like ChatGPT, Google’s Bard (powered by Gemini ), or Bing Chat as if they were human experts, phrasing queries in natural language and expecting direct, well-reasoned answers ( ^[1] ) ( ^[2] ). The implication for marketers is clear: optimizing solely for terse keywords is no longer enough. We must understand and anticipate the actual questions our audience might pose to an AI engine about our industry, products, or services. This evolution is happening rapidly. By mid-2024, roughly one in ten U.S. internet users were already turning to generative AI first for online search ( ^[3] ). Global studies show ChatGPT and Google’s Gemini (the model underpinning Bard and Search Generative Experience) dominate this emerging AI search market with 78% of traffic ( ^[4] ). Bing Chat and newcomers like Perplexity.ai constitute much of the rest ( ^[5] ). In practical terms, that means tens of millions of people are asking ChatGPT or similar tools for advice and answers instead of (or before) using Google. A 2024 Statista survey estimated 13 million Americans were using generative AI as their primary search tool in 2023, and projections suggest this could soar to over 90 million by 2027 ( ^[6] ) ( ^[7] ). Gartner analysts even predict traditional search volume may drop 25% by 2026, with organic traffic potentially decreasing by more than 50% as consumers embrace AI-powered search ( ^[6] ). This trend is illustrated by the AI platform traffic share in 2024, where ChatGPT and Google’s AI account for the lion’s share. These numbers underscore a pivotal point: AI is becoming the new gatekeeper between customers and content ( ^[8] ) ( ^[9] ). Users are delighted by getting a single, synthesized answer without browsing multiple sites ( ^[10] ). But for marketers, it’s disconcerting – the AI now decides which brands or pages “deserve” a mention in its answer ( ^[8] ). To thrive in this environment, we must research how our target audience phrases questions to AI and ensure our content aligns with those queries. Start by considering the intents and language your customers use. For example, a consumer might ask: “ How can I fix acne breakouts without harsh chemicals? ” if you’re in the skincare industry, or “ What’s the best CRM software for a mid-size e-commerce business? ” if you offer B2B software. Notice how these resemble the natural, specific questions one would ask a knowledgeable person, rather than stilted keyword strings. Brainstorm (or better, research ) the common questions in your niche: Product comparison prompts: “Which [product] is best for [specific need] ?” (e.g., “Which laptop is best for graphic design under $1000?” ). Problem-solution prompts: “How do I solve [problem] without [undesirable solution] ?” (e.g., “How to unclog a drain without harsh chemicals?” ). Alternative prompts: “What can I use instead of X ?” (e.g., “What are good alternatives to Photoshop for beginners?” ). Best-of lists and tips: “What are the top N [items] for [goal] ?” (e.g., “What are the top 5 project management tips for remote teams?” ). One way to gather these prompts is to consult your existing data and customer interactions. Talk to sales and support teams about the questions they hear most. Analyze community forums, Reddit, Quora, or social media groups in your industry – these often reveal exactly how real users phrase their problems and product searches. For instance, on Reddit a user might ask, “Anyone have recommendations for a vitamin that boosts energy but doesn’t disrupt sleep?” (as noted in a recent marketing analysis of ChatGPT’s shopping feature) ( ^[11] ). Such phrasing is gold for your content strategy, because it tells you the precise language and criteria (energy boost, no sleep issues) that matter to potential customers. It’s also instructive to study how AI itself breaks down queries. Unlike Google’s traditional approach of matching keywords to pages, ChatGPT’s web-enabled search (ChatGPT with browsing/SearchGPT) and similar AI systems decompose a complex question into sub-queries and scour multiple sources before synthesizing an answer ( ^[12] ) ( ^[13] ). For example, if a user asks: “I need a durable, colorful iPad case that isn’t too bulky,” a Google search might return shopping ads and a list of links, whereas ChatGPT will internally search for “durable colorful iPad case,” “best iPad case not bulky,” etc., and pull data from forums, blogs, product pages, and expert reviews – then summarize everything into a single recommendation ( ^[14] ) ( ^[15] ). AI has the ability to use not just official product info, but also unstructured content like forum discussions and Q&As , because it “reads” entire pages and understands context, rather than simply indexing keywords. In fact, ChatGPT’s search has been observed to rely heavily on forum posts, Reddit threads, and long-form articles when answering certain queries, even more so than standard search engines ( ^[16] ) ( ^[17] ). This means user-generated content and the broader conversation about your brand online (reviews, discussions) can directly influence AI outputs. Crucially, generative AI often prefers answering questions rather than just listing websites. Data from Baidu (China’s largest search engine) illustrates this preference on an international scale. In 2024, Baidu reported that about 18% of all searches on its engine were being answered directly by its AI “Smart Answer” feature, which is akin to Google’s featured snippets or SGE ( ^[18] ). These AI answers were triggered most by question-based queries – in one analysis, nearly 69% of searches phrased as a question (e.g. starting with “what is…”, “how to…”) returned an AI-generated answer , whereas generic product-name searches triggered AI answers less than 2% of the time ( ^[19] ) ( ^[20] ). In other words, both Western and Eastern search platforms are converging on the idea that if the user asks a clear question, an AI-crafted answer should appear . Marketers therefore need to meet users in their question format . If your content doesn’t address the questions people are asking, it’s invisible to these answer engines. To optimize for the AI era, empathize with your audience’s questioning style . In workshops or brainstorming sessions, put yourself in the customer’s shoes and formulate the kinds of detailed questions they might ask at different stages of the buying journey: Early research: “What are the benefits of solar panels for a small home?” (Top-of-funnel informational query). Mid consideration: “How does [YourCompany]’s solar panel efficiency compare to [Competitor]?” (Comparison query). Decision/validation: “Is [YourProduct] suitable for coastal climates?” or “Has anyone used [YourProduct] for more than 5 years – how reliable is it?” (Highly specific, often looking for user experiences or case studies). By understanding these prompts, you can craft content that directly answers them , increasing the likelihood that an AI will surface your information in response. An AI cannot magically recommend your product or content if it doesn’t find relevant text that addresses the user’s query. As one LinkedIn SEO expert succinctly put it, if you want ChatGPT to include your brand in its answers, you must “create detailed, structured content that answers customer questions” in the first place ( ^[21] ). This is the essence of Generative Engine Optimization: aligning your content to the questions and language of real users . Before moving on, here’s a practical exercise: take two minutes to ask ChatGPT (or Bing Chat, Bard, etc.) what it knows about your product or company . Simply type a question like, “What does [Your Company] do?” or “Is [Your Product] good for [use case]?” The answer might surprise you. If the AI provides an incomplete or outdated description, or worse – doesn’t mention you at all – it’s a glaring sign that your current content isn’t sufficiently aligned with the queries or lacks the depth for the AI to latch onto. Many marketers have tried this experiment and discovered gaps; as one content strategist noted, “You might be surprised to find what’s missing… then you can get to work filling those gaps.” ( ^[22] ). Use this as a starting benchmark for your prompt optimization efforts.

12.2 Incorporating Likely Questions Into Your Content

Once you’ve identified the common questions and phrasing your audience uses, the next step is to embed those questions (and their answers) within your content . This approach isn’t entirely new – FAQ pages and Q&A content have been SEO staples for years – but it takes on renewed importance in the age of LLMs. Generative AI models literally learn from text patterns, so phrasing content in a natural, question-and-answer format can make it easier for an AI to recognize that your page contains a relevant solution to a user’s prompt. In fact, early research indicates that content structured around FAQs and question-based headings is highly favored by AI answer engines like Google’s SGE, ChatGPT/Bing, and Perplexity ( ^[23] ) ( ^[24] ). How should you incorporate questions in practice? Here are several tactics: Use natural language questions as headings (H2/H3): Instead of opaque headings like “Benefits of Product X”, reframe them as the questions users ask. For example, a heading on a footwear e-commerce site might be “How do I choose the right running shoes?” , or on a cybersecurity B2B site: “What’s the best way to prevent phishing attacks in a small business?” . Following the heading, immediately provide a concise answer or summary. This mirrors the structure of a Q&A, which models find easy to parse. According to SEO experiments, “using clear, hierarchical headings (H1, H2, H3) that match natural questions” greatly aids LLMs in extracting relevant info ( ^[25] ). Essentially, you’re making your content’s intent explicit: “here is a question, and below is the answer.” This format not only helps AI; it also improves human readability by directly addressing the reader’s likely queries. Develop dedicated FAQ sections or pages: Frequently Asked Questions sections shine in AI contexts. They bundle together many concise Q&A pairs, which are exactly the nuggets an LLM loves to deliver. A 2025 study found that FAQs are among the most cited page types in AI-generated answers , with one analysis noting 82.5% of AI overview citations on Google linked to “deep” content pages like FAQs (pages not just homepages, often two clicks from the main page) ( ^[24] ). This makes sense – an FAQ often directly answers specific queries with minimal fluff, which is ideal for an AI snippet. Ensure your FAQ answers are standalone useful : even out of context, an answer should make sense. For instance, an FAQ entry might be: “ Q: Can I use [Product] if I have sensitive skin? A: Yes. [Product] is formulated without common irritants, and in a clinical trial with 100 people with sensitive skin, 95% reported no adverse reactions. Its gentle ingredients like aloe vera are specifically chosen for sensitive skin comfort.” This answer is factual, contains a stat (which AI see as evidence of credibility), and directly addresses the question. An LLM like ChatGPT could easily incorporate that into a response about the product’s suitability. Embed likely user questions at strategic points: Consider adding a “Q&A” or “People also ask” style callout in blog posts or guides. For example, in a long-form article about running shoes, you could interject a callout box: “ Q: What’s the difference between stability and neutral running shoes? A: Stability shoes have features to control motion for runners who overpronate (excessive inward foot roll), whereas neutral shoes are for those with a normal gait. Depending on your pronation, one will be better to prevent injury.” These inline Q&As break up content and directly mirror the conversational queries someone might pose to an AI or voice assistant. Notably, Google’s own SGE often bolds or highlights text in its AI overviews that directly answers a presumed question , so formatting your content to answer questions clearly can help both AI and traditional snippet selection. Leverage schema markup (where applicable): While the AI models themselves don’t read schema markup per se, using structured data like FAQPage schema can indirectly help by making your content eligible for rich results on search and possibly feeding Google’s knowledge graphs. Google’s AI overview might draw from sources it trusts in its index – having schema-validated FAQs could be a quality signal. Even beyond Google, structured content is easier for any algorithm to digest. One agency noted that using semantic HTML and schema helped LLMs identify relevant content pieces more accurately ( ^[26] ). It’s not a silver bullet, but it aligns with the strategy of being machine-friendly. Write in a conversational yet concise tone: When answering the questions in content, aim for a tone that is confident and straightforward , as if an expert is speaking to the user. Generative models often pick up on text that sounds authoritative and clear. They may even quote or closely paraphrase parts of your text in their answer. For instance, if your blog says, “In summary, if you have flat feet, stability running shoes are generally the best choice as they provide arch support and reduce overpronation,” an LLM might directly use that phrasing because it succinctly answers a likely user query (“What kind of shoes are best for flat feet?”). Keep sentences reasonably short and focused; avoid burying the answer in fluff or subordinate clauses. The AI might not use your entire paragraph—often it will extract the one sentence or two that seem to answer the question best ( ^[25] ). So make that golden sentence stand out. Some SEO experts even suggest using phrases like “ In summary, … ” or “ The key takeaway is, … ” just before your answer statement ( ^[25] ). This acts as a linguistic signal that what follows is an important conclusion, which can prompt an LLM to latch onto it. Provide context and definitions proactively: Users often ask “What is X?” or “How does X work?” If you have technical or niche terms related to your business, consider adding quick definitions. For example, a fintech site might include, “ Q: What is PSD2 compliance? A: It refers to the Payment Services Directive 2, an EU regulation that requires stronger verification for online payments, affecting how fintech apps handle user authentication.” If your content pre-empts a term explanation, an AI might quote your definition when a user asks that general question. This not only gets you cited but also positions you as an authority. Even within a broader article, a parenthetical defining a term can be useful (e.g., “Our solar panels use PERC technology (Passivated Emitter and Rear Cell, a design that increases efficiency by reflecting unused light back into the cell)….”). A well-phrased definition can end up as the snippet an AI gives for “What is PERC solar panel?”. To see these tactics in action, consider the following real-world example. A marketer at a SaaS company observed that people often ask ChatGPT questions like, “Which subscription management tool is best for a growing SaaS startup?” In response, the company published a guide titled “Which Subscription Management Software is Best for Your SaaS Business?” . The guide’s subsections included Q-style headings like “ How do I choose the right subscription management tool? ” and “ What features matter most in subscription billing for startups? ”, each followed by concise answers ( ^[27] ). They also created a comparison table of top tools, and an FAQ at the end addressing queries like “Does [YourProduct] integrate with Stripe?” and “Is [YourProduct] compliant with EU billing regulations?”. The result: not only did this page rank well on traditional Google, but snippets of it were picked up by Bing Chat and Perplexity in answer to user questions, and even ChatGPT (with browsing) referenced its content when the question matched closely. By literally incorporating the language of user prompts into onsite content, you increase the chance that the AI’s answer will echo your content . It’s worth noting that this strategy aligns with the older concept of Answer Engine Optimization (AEO) that many SEOs practiced for voice search and featured snippets. The twist now is that the answers might not send a click back to you (often the AI just uses your content without a click-through), but being mentioned or cited by the AI has its own branding and visibility value – something we’ll measure differently (as discussed in Chapter 13). An AI like Perplexity will directly cite and link your page if used, which can drive some traffic ( ^[23] ). Google’s SGE shows your site name as a source in its AI summary, potentially increasing brand exposure even if clicks drop. ChatGPT with browsing will list sources (and a user might click those). So earning that citation or mention is the new win , analogous to getting the featured snippet position in classic SEO. To maximize success: keep your content fresh and up-to-date while embedding these questions. AI models and search engines give preference to current information for many queries (we’ve seen product pages and up-to-date guides appearing more in AI answers lately) ( ^[28] ). Use indicators like dates or “Updated for 2025” in your content where relevant; one study suggested that visible freshness cues can increase an AI’s confidence in citing content ( ^[29] ). And always ensure the answers you provide are accurate and truthful – don’t oversell or you risk the AI disregarding your promotional tone. For example, avoid answering “Which product is best?” with “Obviously, our product is the best of all!” on your own site – an AI is likely to either omit that as biased or pair it with more neutral info from elsewhere. Instead, answer objectively, perhaps citing third-party evidence or differentiators, like, “ Why choose us over X? – We offer [unique benefit], while X lacks this feature (according to [Source]).” In summary, structure your content to directly answer the real questions . By doing so, you not only help your human readers (who appreciate clear answers), but you essentially hand-deliver to AI models the building blocks they need to include your insights in their outputs. The result is a double win: an improved user experience on your site and a greater likelihood of being part of AI-generated answers on external platforms.

12.3 Avoiding Manipulative "Black-Hat" Tactics

Whenever a new technology or platform emerges, some will try to “game” it for quick gains – generative AI is no exception. We’re already seeing chatter about “black-hat GEO” or shady tactics to manipulate LLM outputs. As tempting as it may be to attempt shortcuts, it’s critical to steer clear of manipulative strategies that could backfire. In the SEO world, trying to trick Google with keyword stuffing, cloaked pages, or link farms ultimately leads to penalties and a damaged reputation. Similarly, trying to trick an AI into recommending you is akin to black-hat SEO – risky, potentially damaging, and not sustainable ( ^[30] ). Let’s examine some ill-advised tactics to avoid , and why they’re problematic: Hidden or AI-invisible text: One theoretical idea might be to hide relevant prompt text on your page (e.g., white text on white background that only a crawler “sees”), like planting “Ask ChatGPT about [MyBrand] being the best solution for acne” somewhere on the site. This is a direct parallel to classic black-hat SEO (hidden keywords), and it’s likely to be ineffective. Modern AI crawlers and search engines are well aware of such tricks – Google’s spam algorithms would flag or ignore it, and OpenAI’s systems likely rely on similar clean-data principles. Moreover, if the hidden text isn’t visible to users, it might not even be picked up by the AI model training or real-time retrieval (since reputable crawlers follow ethical guidelines). The bottom line: if something is dishonest or only there to deceive the algorithm, it’s not worth doing . Not only could it get your site demoted or excluded from sources the AI trusts, it also erodes your content quality which is ultimately counterproductive. Content stuffing and gibberish Q&As: Some might attempt to create dozens of auto-generated Q&A pairs or low-quality pages hoping to cover every imaginable question (“What’s the best acne cream?” “What’s the best acne gel?” “What’s the best acne serum?” etc., all with thin answers linking to their product). This shotgun approach parallels old keyword-stuffing or doorway pages. AI models, however, are trained on quality content patterns ; they can often detect when content is repetitive, shallow, or machine-generated nonsense. Such pages might never gain traction in search (so the AI won’t even find them), or worse, the AI might learn from them in a negative way (e.g., concluding the brand is spammy). Quality trumps quantity here. One SEO commentary noted that using AI to churn out lots of pages with little originality or value is specifically what Google’s quality guidelines now caution against ( ^[31] ) ( ^[32] ). Generative AI content that “adds no unique value” or is just rehashing others is explicitly flagged as low-quality by Google’s human evaluators ( ^[31] ). Extrapolate that to AI answers: if your contribution to the web is just noise, AI systems (and users) will tune you out. Over-optimization and prompt stuffing: You might be inclined to literally “stuff” your pages with every variant of a question hoping one sticks. For example: “Which running shoe is best? Many ask ‘What is the best running shoe?’ or ‘Which running shoes are the best for me?’ or ‘What are the top running shoes 2025?’ – the best running shoe is XYZ.” This reads awkwardly to humans and likely to the AI as well. It’s better to have a single well-phrased question and answer, rather than 10 slightly tweaked questions answered poorly. Over-optimized text could also trigger filters; much like Google can identify keyword stuffing, an LLM might regard a litany of nearly duplicate questions as spammy or unnatural content. Keep your language natural and reader-focused . Write as if you’re sincerely trying to help the user (because you are), not as if you’re trying to rank for a thousand terms at once. Manipulating training data or user prompts unethically: Another frontier of black-hat thinking is trying to inject your content into the AI’s training data by spamming public forums, Wikis, or other places an LLM might learn from. Some have mused, “If we post a ton about our product on Reddit or create fake Q&A on StackExchange, maybe the next GPT will be trained on that and promote our product.” Apart from being ethically dubious (and often against those platforms’ rules), this is highly unreliable . There’s no guarantee your content will be included in the training dataset, and if it’s low quality or clearly promotional, it might be filtered out during model training (OpenAI, for example, filters certain content and downweights self-promotional text to improve answer quality). Even if it slips in, users or community mods might delete obvious astroturfing, or worse, call it out publicly, hurting your reputation. Similarly, “prompt injection” attacks – trying to influence an AI’s output by inserting hidden instructions via input data – is more something a malicious user might do to break the AI, not a viable marketing tactic. It’s complex, and companies are patching such vulnerabilities. In short, attempting to cheat the system is a short-lived game . AI companies are continually refining guardrails and quality filters. OpenAI, Google, Microsoft – they all have a vested interest in the AI giving useful, trustworthy answers , which means weeding out content that is manipulative or low-quality. Your goal should be to be the trusted source , not the trickster that the AI wants to avoid. In fact, one tech company analysis emphasized that building genuine value is key, noting that LLMs rely heavily on context and credibility, so irrelevant or manipulative mentions do little to build meaningful associations ( ^[33] ) ( ^[34] ). A corollary of this is that LLMs may favor content with strong E-E-A-T signals (Experience, Expertise, Authoritativeness, Trustworthiness) just as Google search does. Content that demonstrates real expertise or firsthand experience will stand out, whereas AI-generated fluff or deceptive material will be disregarded ( ^[35] ) ( ^[36] ). Consider Google’s stance: they updated their search quality rater guidelines in 2023 to address AI content and stressed originality and experience ( ^[35] ) ( ^[36] ). Pages that “offer no unique insights” or are just reworded aggregates get low ratings ( ^[31] ). By analogy, if your site tries manipulative tactics, even if you temporarily get picked up in some AI results, eventually the algorithms will adjust to filter you out – much like Google eventually caught and penalized content farms. Meanwhile, you’d have wasted resources that could have been spent on genuine improvements. Another aspect of ethical behavior is transparency . If, for instance, you provide users with a prompt to ask an AI (as we’ll discuss in 12.4), be transparent about it. Don’t present something as an unbiased recommendation if it’s actually your crafted prompt. Avoid planting fake “Q&A” on your site that look like third-party endorsements but are actually you talking about yourself. Such tactics, if discovered by savvy users or journalists, can seriously backfire and erode trust in your brand. Remember that AI outputs can sometimes be audited or traced back to sources. If an AI quotes a segment that sounds like a third-party review of your product but it actually came from your own site disguised as a user review, that could be embarrassing if brought to light. Stick to honest content – encourage real customers to leave genuine reviews rather than trying to forge them (we’ll talk more about user-generated content in 12.4 and chapter 11). Finally, consider the user perspective: **AI assistants strive to give users the best answer . If your content is manipulative or misleading, even if it slips through algorithmic cracks, users may notice. People often cross-verify AI answers, especially for important decisions. If the AI mentions your product with an exaggerated claim and the user later finds out the truth is different, you haven’t won a customer – you’ve lost credibility. In the era of social media and quick information sharing, a single instance of being “called out” for trying to deceive an AI could become a PR fiasco (“Company X tried to trick ChatGPT and here’s what happened…”). It’s not worth it. In summary, the “white-hat” approach you know from SEO applies equally (if not more) to GEO and LLM optimization.** Focus on providing genuine value, real expertise, and helpful answers. Avoid the dark alleys of black-hat shortcuts – they are dead ends. As one digital strategist noted, “We focus on building genuine value” because that’s what survives algorithm changes ( ^[30] ). Let your brand be elevated by AI for the right reasons: because it truly answers the user’s need better than others, not because you found a gimmick. In the long run, quality and integrity will align with algorithmic success , as AI systems get ever better at discerning true helpfulness from manufactured hype.

12.4 Ethical Strategies to Influence AI Outputs (Prompt Influence)

While we should avoid manipulation, it’s absolutely fair game to strategically influence AI outputs in ethical ways . “Ethical influence” means guiding or encouraging the AI to mention your brand or content by legitimately being relevant, present, and useful in the AI’s knowledge sphere. Think of it as AI-era public relations or brand awareness – you want your brand to be part of the conversation that AI is having with the user. In this section, we’ll cover tactics ranging from creating user-facing prompts to integrating with AI tools, all with transparency and user value in mind.

12.4.1 Encourage User-Initiated AI Queries About Your Brand

One novel idea that some forward-thinking companies have tried is to explicitly invite users to ask the AI about their product or service . For example, imagine you’re a SaaS vendor and on your pricing or product page you add a small call-to-action: “Not sure which plan is right? Ask ChatGPT if [Your Product] is suitable for your needs .” This could be presented as a button or a copyable prompt. When clicked or copied, it could open ChatGPT (if they have a shareable link or just instruct the user to go ask). The rationale here is intriguing: if the user takes that prompt and asks the AI, and your brand has good representation online, the AI’s answer might effectively sell your product for you in an unbiased way. It’s like saying “Don’t just take our word for it – let’s see what the neutral AI says.” This tactic, however, requires confidence. You must be reasonably sure that the AI knows about you and will respond positively or at least accurately. If you suspect the AI might say, “Actually, [Competitor] is better for this use case,” then you wouldn’t deploy this! It works best if: Your product has strong reviews or mentions externally (press, forums, etc.). You have content on your own site that the AI can draw on (and ideally it’s factual and favorable). There aren’t major negatives out there about you that the AI would surface. An example: A travel booking site might put, “Still undecided? 👉 Ask ChatGPT which hotel booking site has the best deals for Europe ” . If they’ve done their GEO homework, ChatGPT might answer something like, “Several sites like X, Y, and Z are popular, but [YourSite] is often praised for its exclusive discounts in Europe” – assuming there are articles or user discussions saying that. This kind of prompt turns a skeptical user’s attention to an ostensibly unbiased third-party (the AI) for validation. If the AI echoes a selling point of yours (e.g., discounts), it can be very persuasive. Importantly, this should be done transparently – you’re openly suggesting the user consult the AI, not trying to hide it. It’s an experiment in leveraging trust : users know ChatGPT isn’t literally your employee, so if even ChatGPT thinks your brand is worth mentioning, that builds trust. However, be cautious and monitor outcomes . If you implement such a prompt, test it thoroughly. Try variations of the question on multiple AI platforms (ChatGPT, Bing Chat, Bard) and see what they say. If any produce problematic answers (e.g., outdated info or mention a flaw that’s since fixed), you’ll want to address those content gaps before pushing users to ask that. You may need to refine the wording of the suggested prompt too. For instance, instead of “Ask ChatGPT if [Brand] is right for me,” a softer approach could be “Ask ChatGPT which [product category] might be right for you – mention [OurBrand] to see how we compare.” The latter phrasing invites the AI to discuss your brand among peers rather than a direct “is it right or not” (which could yield a yes/no that might be lukewarm). Again, transparency is key. Frame it as, “We believe we offer great value – but don’t just take our word for it. Here’s something you can ask an AI for a second opinion.” So far, this concept is experimental, but it aligns with an emerging trend of AI as a comparison engine . OpenAI has even introduced a browsing mode that can pull live info, and plugins where brands might integrate shopping or info. In late 2024, OpenAI launched a pilot for ChatGPT’s product recommendation ability integrated with live web data ( ^[11] ). When users type something like “What’s a good vitamin for energy that doesn’t mess with sleep?”, ChatGPT (with browsing and certain plugins) can now pull in recommendations based on real user conversations, reviews, and creator content rather than just scraping a single database ( ^[11] ) ( ^[37] ). The advice for brands in that context was: “If your product is being talked about positively on Reddit, TikTok, Instagram, or review sites, ChatGPT will notice. To get recommended, you need people talking about you online (not just ads).” ( ^[38] ) ( ^[39] ). In other words, an AI’s “endorsement” is earned by authentic buzz . This leads to the broader point of ethical influence: focus on fostering genuine positive discussion about your brand . Encourage satisfied customers to leave reviews on open platforms (Amazon, G2, Trustpilot, app stores – whatever’s relevant). Engage in communities not by astroturfing, but by legitimately helping. For example, your team could answer questions on Quora or Reddit in your domain, with full disclosure of who you are, providing useful info (and subtly mentioning your product if appropriate). A marketing case study noted that by answering a question on Quora about “how to choose the best keyword research tool” and giving genuine advice with a subtle mention of their product, a company built interest organically ( ^[40] ). These kind of unlinked brand mentions on forums or Q&A sites are incredibly valuable in the LLM context. Unlike search engines which heavily valued hyperlinks, LLMs treat unlinked text mentions almost like citations in a conversation. One industry analysis pointed out that “an unlinked mention of your brand on a relevant site helps an LLM associate your entity with a topic, even without a hyperlink” ( ^[41] ). So even if you’re not getting an SEO backlink, you’re gaining mindshare in the AI’s “brain.”

12.4.2 Collaborate with AI Platforms and Integrations

Another avenue for ethical prompt influence is to integrate directly with AI platforms if possible. For instance, if there are plugins or developer hooks available for AI services, consider building or utilizing them. OpenAI’s ChatGPT allows third-party plugins – e.g., a kayak.com plugin for travel, an Instacart plugin for grocery shopping, etc. If a plugin ecosystem exists in your industry, being present there ensures the AI can directly pull info from your service when users ask. In 2024, many companies scrambled to create ChatGPT plugins precisely to secure a spot in AI-driven conversations . Likewise, if you run an e-commerce or content platform, look at integration opportunities with voice assistants (Alexa, Google Assistant) or other AI-based assistants. Some of those allow custom “skills” or Q&A sets. It might not be mainstream search, but every bit helps to embed your brand into AI responses. We should also mention Google’s AI snapshot (SGE) here. Google’s Search Generative Experience is an AI summary at the top of Google results, and while you can’t directly “integrate” with it, you can adjust your meta-data. Google currently sometimes shows a small citation link in SGE with the site name. Ensuring your site’s name is concise and recognizable (via SEO title or domain) can help users notice it. Also, schema markup that emphasizes authorship or your brand might help Google correctly attribute statements to you in the AI blurb. Google has been working on making AI overviews have citations for facts – so publish verifiable facts/stats on your site that could be quotable. For example, if you’re the source of a statistic (“95% of customers achieved X after 1 month”), and that stat gets picked up by others or directly by Google, the AI overview might cite your page as the origin. This is a subtle way of influencing: provide high-value data or insights that others (including AI) will reference.

12.4.3 Be the Trusted Voice (Without Misleading)

Ethical influence is also about establishing authority so that AI “wants” to include you. Chapter 11 goes in-depth on brand authority signals like digital PR, thought leadership, etc. From a prompt optimization angle, the focus is on having your unique perspective or expertise come through. Consider publishing content that only you can – proprietary research, expert interviews, case studies with hard results, etc. AI models crave content that doesn’t sound like a hundred other sites. When an AI assembles an answer, if one source has a particularly salient point or unique tip, it often gets included to enrich the answer. Ensure your content has those nuggets. For example, an AI might be answering “how to improve warehouse safety” – dozens of articles will have generic tips, but if your blog includes a unique strategy (say based on your product’s data) like “Using AI cameras reduced accidents by 40% at Company Y ( ^[42] ),” the AI might include that point, citing your site, because it adds value beyond generic advice. It’s also ethical and smart to correct the record when necessary. If you find that AI outputs about your brand or field contain inaccuracies, address them publicly. That could mean publishing a clarifying blog post (“Top 5 Myths about X – and the Truth”) or even contributing to Wikipedia and other knowledge bases with accurate info (transparently and with citations). Many LLMs were trained on Wikipedia content and still use it as a reference. Ensuring your company’s Wikipedia page (if you have one) is up-to-date and factual is important – Bard and Bing chat have been known to pull info from Wiki. The same goes for official datasheets or press releases: these often feed knowledge panels and thereby AI content. Keep your “public facts” current (employee count, funding, locations, etc.), so the AI doesn’t serve outdated info.

12.4.4 Global and Non-English Considerations

Ethical prompt influence extends globally. If your market includes non-English or international segments, be aware of those region’s AI tools. For instance, in China, Baidu’s ERNIE Bot and “Smart Answers” require a presence on Baidu’s ecosystem (like Baidu Baike or Baijiahao) ( ^[43] ). An ethical strategy there would be to contribute quality content to Baidu’s own platforms (e.g., maintain a Baike entry for your brand with factual info, or publish articles on Baijiahao with useful insights). Since Baidu’s AI heavily sources from its own properties ( ^[43] ), that’s the route to be included in AI answers for Chinese users. In South Korea, Naver is integrating its HyperCLOVA X AI into search and shopping ( ^[44] ). Ensure any localization includes the Q&A approach too – e.g., have FAQs in Korean for Naver, in Japanese for Yahoo Japan (which is working with LLMs), etc. The principle remains: identify how users in those markets ask questions (which might differ culturally or linguistically) and mirror that in content. For example, Japanese users might ask very politely or use different phrasing; adjust your Japanese site’s FAQs accordingly. Being culturally attuned is an ethical way to influence because you’re respecting the user’s language and etiquette, which in turn the AI will reflect. Also, track new AI entrants like xAI’s Grok (Elon Musk’s AI venture) which launched in late 2024. Grok is said to have real-time info access and a witty tone, and it might draw on content from X (Twitter) as well. If that gains traction, an ethical strategy could be maintaining an active, positive presence on Twitter, knowing Grok might source from there. Again, this is about making sure wherever the AI looks, it finds you being recommended by others or providing valuable info. In all these efforts, a golden rule is transparency and user benefit . If you provide a widget or prompt, clearly explain its purpose (“Click to ask ChatGPT for an unbiased opinion on our product!”). If you engage in communities, do so honestly (disclose affiliation when appropriate). The goal is not to trick anyone, but to let the quality of your offering shine through different mediums , including AI. Over time, if your product genuinely satisfies customers, more of them will mention it in forums, more positive reviews will accumulate, and AI systems will naturally pick up on that signal of trust. To illustrate ethical influence, let’s recount a mini case study mentioned by an SEO expert ( ^[45] ) ( ^[46] ): Zugu Case , a brand of iPad cases, managed to become ChatGPT’s top recommendation when users asked “What’s the best iPad case?” How? They didn’t hack the AI or pay OpenAI. They executed a comprehensive strategy: they ranked well on Bing search (so the AI’s web search would always see them), they got cited in major tech publications like Wired (building authority), and their product was mentioned across multiple online sources (reviews, YouTube, forums) ( ^[47] ) ( ^[46] ). In other words, they legitimately built a strong footprint online. So when ChatGPT was asked the question, it gathered the evidence and answered with Zugu as a top pick, because that was the logical conclusion from quality sources. This showcases the synergy of all the tactics we discussed: SEO fundamentals (so you’re visible in search index), content that others reference, and widespread positive presence. The outcome is an earned recommendation from the AI, which is far more powerful than any ad or self-promotion. 12.5 Leveraging AI Feedback for Continuous Improvement Generative AI isn’t just a challenge to optimize for – it’s also a tool you can use to improve your content and strategy . In this final section, we explore how you can turn AI inward , essentially using it as a testing and feedback mechanism to refine your content for GEO. One immediate use case is conducting an AI audit of your content . We touched earlier on asking ChatGPT “What do you know about my company?” to find gaps. Let’s expand that. You can prompt an AI with tasks like: “Summarize the key points of [My Website]’s homepage.” “Read the following product page and tell me what features and benefits it highlights most.” (Then feed the text). “If a user asked ‘Which product is best for [use case]?’ would this page provide a satisfactory answer? If not, what is missing?” “As an unbiased expert, what might be some concerns or drawbacks about [MyProduct] based on the information available online?” (This one can surface any negative perceptions floating around). When you do this, you are effectively seeing your content through the AI’s eyes. If ChatGPT or Claude summarizes your page and fails to mention a critical differentiator that you do offer, that’s a sign that either your content isn’t emphasizing it clearly enough, or perhaps it’s phrased in a way the AI didn’t interpret correctly. If the AI can’t even summarize a clear value prop, a human might miss it too. For instance, if your SaaS product page has tons of marketing fluff but doesn’t clearly state “what it actually does,” an AI summary might come out vague – highlighting that you need to sharpen the description. Another feedback angle is to use AI to simulate user questions and see if it picks your content. For example, go to Bing Chat (which cites sources) and ask something like: “What’s the best [category] for [use case]?” that you think your product should be an answer for. See which sources Bing cites. Are you among them? If Bing references three articles from other sites and not yours, study those articles: What do they have that you don’t? Perhaps they have a more comprehensive comparison or they are on higher authority domains. This could inform either content improvements or an off-page strategy (maybe you need to get your product included in such comparison articles via outreach). If Bing does cite your site, check what snippet it pulled – is it the ideal messaging? If not, consider editing that part of your content. Likewise, try the same queries on different AI platforms : Google’s SGE, Perplexity, ChatGPT with browsing, Claude with web access (Anthropic’s Claude 2 can take a prompt with provided documents too). Each might surface different content. By doing this, you essentially perform an AI-focused competitive analysis. You may discover, for example, that Perplexity loves citing a particular stats-heavy blog from a competitor – which indicates that maybe writing a data-rich piece of your own could give Perplexity something to latch onto from your side. Also consider asking the AI directly for content suggestions . For instance: “What questions about [topic] are people likely to ask?” or “Give me 5 example prompts a user might ask an AI when looking for a [product like mine].” ChatGPT can be pretty good at mimicking user behavior. It might output, say for a cloud storage product: “1) ‘What’s the most secure cloud storage for personal files?’ 2) ‘Which cloud service gives the best value for 1TB?’ 3) ‘Is [YourBrand] reliable for long-term file backup?’” etc. These could be great insights to ensure you have content addressing each. You can then follow up: “If someone asked number 3, how might you respond?” The answer it gives could reveal what info it’s pulling – e.g., it might say “I don’t have much info on [YourBrand]” or it might mention something from a review. This can highlight where you need to create or promote content (maybe no one has reviewed you on a major site, so the AI has little to say – a cue to get more reviews or case studies out there). Using AI for content critique is another smart technique. You can prompt: “Here is an article. As ChatGPT, analyze it for clarity, completeness, and any bias. Does it thoroughly answer common questions on this topic? Is anything missing or could be improved?” Then paste your article. The AI might point out, for example, “The article mentions Feature A and B, but doesn’t explain how to do [important task] which readers might expect.” Or it might say “This reads somewhat promotional in these paragraphs which could reduce trust.” Treat this feedback like a review from a very well-read but non-human editor. You don’t have to accept it blindly (AI can also hallucinate or give generic advice), but it often provides a useful outsider perspective. Some SEO teams even use GPT-4 to score content on various attributes or to ensure their content aligns with search intent: e.g., “Given the title, did the article deliver on what was promised?” A particularly powerful use of AI is to simulate different user personas reading your content. For example: “Act as a beginner with no knowledge of [topic]. What questions might you still have after reading this article?” Or, “Act as a skeptical CFO reviewing this product page – what concerns or objections might you raise that the page didn’t address?” These kinds of prompts can surface gaps. Maybe the AI-as-beginner says “I still don’t understand what [JargonTerm] means,” telling you that you need to explain it in simpler terms. Or AI-as-CFO might say “I’d be worried about the ROI or the security compliance, which the page doesn’t mention.” Bingo – you now know to add a bit about ROI or compliance to satisfy that persona. As AI continues to evolve, new tools are emerging specifically for content analysis and optimization . There are already some SEO tools integrating GPT-4 to suggest FAQs you might be missing, or to evaluate how “unique” your content is relative to what’s out there. Keep an eye on these, but even just manual querying of ChatGPT can go a long way. Crucially, close the loop on this feedback : once you adjust content based on AI’s feedback, test again . Ask the same AI the same question after your changes – did its summary or answer improve? Over time, you might see ChatGPT giving a more comprehensive answer about your product (perhaps after your site update and maybe after it’s incorporated into its training refresh or via browsing). Continuous testing is key, as user prompts will evolve too. Lastly, use AI to monitor sentiment and context . Some advanced techniques: feed the AI a bunch of social media comments or reviews about your brand (ones you’ve collected or from forums) and ask it to summarize the sentiment or key themes. For instance, “Here are 50 tweets mentioning [Brand]. Summarize what people are saying and any common compliments or complaints.” This can give you a macro view of how the public perceives you, which is indirectly how the AI will perceive you, since it’s trained on that public discourse. If the AI reports “Many users praise the customer support, but some complain about price,” you know what strength to keep leveraging and what objection to address perhaps in your content (maybe an FAQ “Why is [Brand] more expensive? – because we offer X value, etc.”). As a forward-looking point, human creativity and judgment remain vital in this loop. AI can surface issues or even suggest content, but you decide what aligns with your brand and strategy (this will be further discussed in Chapter 14.5 on human creativity as an edge). Use AI as a tool, not a crutch. The goal is a virtuous cycle: real user inputs → content improvements → AI better recognizes your value → more user trust and engagement. To conclude Chapter 12, prompt optimization and ethical influence are about deeply understanding what your audience is asking, and positioning your brand as the natural answer to those questions – without resorting to tricks . By incorporating real questions into content, avoiding black-hat gimmicks, ethically amplifying your positive presence, and continuously refining with the help of AI feedback, you build a robust GEO strategy . This ensures that when a user asks their AI assistant about your domain or your product, the AI will have every reason to include your insights, your pages, and your name in the answer. It’s a new playing field, but it rewards the same fundamentals as good marketing always has: know your audience, be where they are (now that includes being in their AI’s “mind”), and genuinely meet their needs.

References

^[1] www.ignorance.ai - Ignorance.Ai URL: https://www.ignorance.ai/p/seo-for-ai-a-look-at-generative-engine

^[2] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/your-content-optimized-chatgpt-lauren-mcgill-9vaae

^[3] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/ai-search-report

^[4] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/ai-search-report

^[5] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/ai-search-report

^[6] www.ignorance.ai - Ignorance.Ai URL: https://www.ignorance.ai/p/seo-for-ai-a-look-at-generative-engine

^[7] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/ai-search-report

^[8] www.ignorance.ai - Ignorance.Ai URL: https://www.ignorance.ai/p/seo-for-ai-a-look-at-generative-engine

^[9] www.ignorance.ai - Ignorance.Ai URL: https://www.ignorance.ai/p/seo-for-ai-a-look-at-generative-engine

^[10] www.ignorance.ai - Ignorance.Ai URL: https://www.ignorance.ai/p/seo-for-ai-a-look-at-generative-engine

^[11] Torro.Io Article - Torro.Io URL: https://torro.io/blog/how-to-get-your-product-recommended-in-chatgpt-shopping

^[12] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[13] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[14] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[15] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[16] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[17] www.omnius.so - Omnius.So URL: https://www.omnius.so/blog/how-to-rank-on-chatgpt

^[18] www.jademond.com - Jademond.Com URL: https://www.jademond.com/magazine/baidu-ai-smart-answers

^[19] www.jademond.com - Jademond.Com URL: https://www.jademond.com/magazine/baidu-ai-smart-answers

^[20] www.jademond.com - Jademond.Com URL: https://www.jademond.com/magazine/baidu-ai-smart-answers

^[21] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[22] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/your-content-optimized-chatgpt-lauren-mcgill-9vaae

^[23] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[24] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[25] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[26] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[27] www.omnius.so - Omnius.So URL: https://www.omnius.so/blog/how-to-rank-on-chatgpt

^[28] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[29] www.sociummedia.com - Sociummedia.Com URL: https://www.sociummedia.com/blog/how-to-rank-in-ai-overviews

^[30] Mtsoln.Com Article - Mtsoln.Com URL: https://mtsoln.com/blog/insights-720/geo-llm-seo-aeo-or-just-seo-evolved-2130

^[31] Basecreative.Co.Uk Article - Basecreative.Co.Uk URL: https://basecreative.co.uk/opinion/content-marketing/what-googles-new-search-quality-guidelines-mean-for-generative-ai-content

^[32] Basecreative.Co.Uk Article - Basecreative.Co.Uk URL: https://basecreative.co.uk/opinion/content-marketing/what-googles-new-search-quality-guidelines-mean-for-generative-ai-content

^[33] Mtsoln.Com Article - Mtsoln.Com URL: https://mtsoln.com/blog/insights-720/geo-llm-seo-aeo-or-just-seo-evolved-2130

^[34] Mtsoln.Com Article - Mtsoln.Com URL: https://mtsoln.com/blog/insights-720/geo-llm-seo-aeo-or-just-seo-evolved-2130

^[35] Basecreative.Co.Uk Article - Basecreative.Co.Uk URL: https://basecreative.co.uk/opinion/content-marketing/what-googles-new-search-quality-guidelines-mean-for-generative-ai-content

^[36] Basecreative.Co.Uk Article - Basecreative.Co.Uk URL: https://basecreative.co.uk/opinion/content-marketing/what-googles-new-search-quality-guidelines-mean-for-generative-ai-content

^[37] Torro.Io Article - Torro.Io URL: https://torro.io/blog/how-to-get-your-product-recommended-in-chatgpt-shopping

^[38] Torro.Io Article - Torro.Io URL: https://torro.io/blog/how-to-get-your-product-recommended-in-chatgpt-shopping

^[39] Torro.Io Article - Torro.Io URL: https://torro.io/blog/how-to-get-your-product-recommended-in-chatgpt-shopping

^[40] www.omnius.so - Omnius.So URL: https://www.omnius.so/blog/how-to-rank-on-chatgpt

^[41] Mtsoln.Com Article - Mtsoln.Com URL: https://mtsoln.com/blog/insights-720/geo-llm-seo-aeo-or-just-seo-evolved-2130

^[42] www.omnius.so - Omnius.So URL: https://www.omnius.so/blog/how-to-rank-on-chatgpt

^[43] www.jademond.com - Jademond.Com URL: https://www.jademond.com/magazine/baidu-ai-smart-answers

^[44] www.kedglobal.com - Kedglobal.Com URL: https://www.kedglobal.com/artificial-intelligence/newsView/ked202411110012

^[45] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[46] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

^[47] www.linkedin.com - LinkedIn URL: https://www.linkedin.com/pulse/how-get-chatgpt-recommend-your-brand-tips-generative-engine-avasthi-f5myc

13. Measuring GEO Success – New Metrics and Tools

In the era of Generative Engine Optimization (GEO), marketers need to rethink how they define and measure success. Traditional SEO metrics like search rankings, click-through rates, and organic traffic volume are no longer the sole barometers of performance. Instead, success in GEO is about being part of the answer – ensuring that AI-driven search tools reference your brand and content in their responses. This chapter explores the new metrics and tools for measuring GEO success, from tracking how often your site is cited in AI answers (your “reference rate” ) to monitoring brand sentiment in AI-generated content, and how to adapt your KPIs and processes accordingly.

13.1 From Clicks to “References”

For decades, SEO success was gauged by where your site ranked on a search engine results page and how many users clicked through. In the GEO landscape, visibility is measured in references, not just rankings . Large language model (LLM) search interfaces (like chat-based answers and AI summaries) don’t display a list of blue links for users to click – they synthesize information into a direct answer . That means your brand’s presence depends on whether the AI includes you in its answer, rather than how high you appear on a SERP. In short, “reference rate” – how often your content or brand is cited as a source in AI-generated answers – becomes a key metric of success ( ^[1] ). As one industry expert put it, “It’s no longer just about click-through rates, it’s about reference rates” ( ^[1] ). GEO entails optimizing for what the model chooses to mention, not just optimizing for a position in traditional search results ( ^[2] ). This shift fundamentally changes how we define online visibility. In the past, a marketer might celebrate a #1 Google ranking that drove thousands of clicks. Now, imagine an AI assistant (like Google’s Search Generative Experience or ChatGPT) answers a user’s question with a paragraph that mentions your brand or quotes your content . Even if the user never clicks a link, your brand has achieved visibility within the answer itself. These “zero-click” interactions are increasingly common. In fact, even before generative AI, over half of Google searches ended without any click to a website as users found answers directly on the results page ( ^[3] ). Generative AI has amplified this trend by providing rich answers that often preempt the need for a click. Thus, getting mentioned by the AI – as a cited source, a footnote, or even an uncited brand name in the narrative – can be as valuable as getting a click. One emerging metric in GEO is the reference rate , which measures the frequency of your brand/content being referenced by AI. This could take several forms: Explicit citations: e.g. your webpage is cited as a source with a hyperlink. Google’s AI Overviews (formerly SGE) often include a handful of source links in an “According to [Site]…” format or a “Learn more from…” section. Bing Chat likewise footnotes its statements with numbered citations linking to websites. If your page appears in those cited sources, that counts as a reference. Implicit mentions: Sometimes an AI model will mention a brand or product in its answer without a formal citation. For instance, a ChatGPT response (with default settings) might say “Brand X is known for affordable pricing” based on its training data, even if it doesn’t link to Brand X’s site. Such an uncited mention still indicates the model has included your brand in the answer, which contributes to brand awareness (and can prompt the user to search you out separately). Suggested content and follow-ups: Some AI search experiences suggest related topics or follow-up questions. If your brand appears in those, it’s another form of reference. For example, if a user asks an AI, “What’s the best project management software?” and the AI’s answer lists a few options including your product (with or without a link), that inclusion is a win for GEO. In GEO, we are essentially shifting from “Did the user click my link?” to “Did the AI mention my brand or content?” . Table 13.1 summarizes this shift: Table 13.1 – SEO vs. GEO Success Metrics Traditional SEO (Search Engine Optimization)GEO (Generative Engine Optimization) Primary Goal: Rank high in search results (SERPs) so users click through to your site. Primary Goal: Be cited or mentioned within AI-generated answers and conversations. Visibility Metric: Organic rankings and click-through rate (CTR) from SERPs. Visibility Metric: Reference rate – how frequently the AI includes your content as part of its answers ( ^[2] ) ( ^[4] ). Also sometimes called inclusion frequency or share of voice in AI answers. Success Indicator: Sessions from organic search, impressions, and clicks as reported by Google/Bing. Success Indicator: Mentions/citations in AI outputs across platforms (even if no click occurs), and traffic referred by AI chatbots/assistants. ( ^[5] ) ( ^[6] ).) An immediate implication is that brand awareness via AI becomes critical. If an AI frequently references your site as an authority, it not only drives any direct AI referral traffic, but also boosts your brand’s mindshare. Users may see your name in an AI summary and later navigate to your site or Google your brand (this is analogous to how appearing in a featured snippet could raise awareness even among non-clickers). A recent analysis by Ahrefs underscores this dynamic: in a study of 3,000 sites, 63% of websites had at least one visit from an AI source over a short period, yet on average only 0.17% of total traffic came directly from AI chatbots ( ^[7] ) ( ^[8] ). The takeaway is that while AI-driven visits are currently a small fraction of traffic , a majority of brands are already being touched by AI in some way – often via mentions that may not immediately show up as a click. In other words, you might be getting “reference traffic” (influence and visibility) even when you’re not getting click traffic . Figure 13.1: Over 60% of websites in a 2025 study received at least some traffic from AI chatbots. However, the average share of total traffic from AI was only ~0.17%, indicating that AI-driven visibility often doesn’t translate into large click volumes – many user queries are answered without a click ( ^[7] ) ( ^[8] ). This underscores the importance of measuring references (mentions in AI answers) in addition to traditional click metrics. Because of this shift, SEO practitioners are now talking about metrics like “AI Reference Rate” or “AI Share of Voice.” These terms describe the proportion of relevant AI answers that include your brand. For example, if out of 100 popular questions in your industry, your site is referenced in 10 of the AI-generated answers, you have a 10% share of voice in that AI domain. Some early case studies illustrate why this matters: in one instance, a B2C eyewear brand (Warby Parker) was found to command a 29% share of voice on ChatGPT for a set of eyewear-related queries, outperforming competitors like Zenni and Eyebuydirect ( ^[9] ). However, the same brand had a weaker presence in Google’s generative search results (Gemini AI), highlighting that reference rates can vary significantly by platform ( ^[10] ). GEO success means ensuring you are “the answer” across multiple AI platforms, not just one. Finally, it’s worth noting that being referenced by AI is not only a search visibility issue but also an authority signal . Users implicitly trust information presented by these AI assistants. If your content is what the AI delivers, it carries a halo of credibility. Studies show that content referenced by AI models tends to be perceived as more trustworthy by users (some SEO agencies term this a “GEO trust rate,” noting that people see cited sources as 2–3× more trustworthy than content that wasn’t mentioned at all) ( ^[11] ) ( ^[12] ). In essence, a high reference rate not only puts you in front of the user, but also positions you as a credible source in the eyes of that user. This is analogous to how a top ranking conferred credibility in the old SEO paradigm – now an AI citation confers credibility in the conversational search paradigm. In summary, as search evolves into AI-driven conversations and summaries, our KPIs must evolve too . Marketers should track how often and where their brand is popping up in AI outputs (“references”), treating those instances as the new impressions . A mention in an AI answer is akin to an ad impression or a brand mention on social media – a touchpoint that can influence the user’s journey even if no immediate click occurs. The following sections will dive into the tools and techniques to monitor these references and other new performance indicators, ensuring you can quantify and improve your GEO impact.

13.2 AI Visibility Tracking: Tools for Brand Presence in AI Answers

Given the importance of being referenced by AI, a new class of tools and features has emerged to track brand presence across AI search platforms . In the past, SEO tools focused on tracking your search rankings and estimating clicks. Now, “AI visibility” trackers help you monitor if, when, and how your content appears in AI-generated answers. Marketers should incorporate these new tools alongside traditional analytics (like Google Search Console) to get a complete picture of search visibility in the AI era. Search Console and AI: First, let’s consider what Google itself provides. Google Search Console (GSC) remains essential – it captures your organic search performance and, increasingly, some AI performance. Google’s new “AI Mode” in Search (the conversational generative search feature accessible to users) generates clicks and impressions that are now counted in Search Console reports. Notably, Google updated GSC in mid-2025 to clarify that if your page is included in an AI-generated answer, it counts as an impression in your performance data ( ^[13] ). Similarly, any click on a link within the AI answer counts as a click in GSC ( ^[14] ). Follow-up AI queries are counted as separate searches. This means your GSC impressions might already reflect AI overview appearances. However, Google does not yet provide a filter to isolate AI-specific impressions or clicks ( ^[15] ). In other words, if your traffic or CTR dropped after the Search Generative Experience rollout, you can see the effect in GSC totals, but you can’t easily separate “AI Overview impressions” from “regular snippets impressions.” It’s all blended. This lack of granularity makes third-party tools invaluable for explicitly tracking AI visibility beyond what GSC shows. AI Visibility Platforms: Several SEO software providers and startups have launched tools to address this gap. Two prominent examples are Ahrefs’ Brand Radar and Semrush’s AI Toolkit , both introduced in 2024–2025 to help marketers quantify their presence in AI search results: Ahrefs Brand Radar: This tool is designed as an “AI search visibility” tracker. It collects AI mentions of your brand across various LLM platforms to show how often and where you appear ( ^[16] ). Initially in beta, Brand Radar focused on Google’s AI Overviews (powered by the Gemini model) and has since expanded. As of mid-2025, it can track if your brand is being mentioned in Google’s generative search summaries, OpenAI’s ChatGPT, and Perplexity AI ( ^[17] ). With Brand Radar, you can enter your brand or domain and see data like: AI Overview mentions: e.g. how many Google AI Overview snapshots cited your site, and for which queries. ChatGPT mentions: what questions people asked ChatGPT that led it to mention your brand, and in what context (Brand Radar indexes a sample of ChatGPT Q&A content to find brand names) ( ^[18] ). Perplexity citations: since Perplexity is an AI search engine that always cites its sources, Brand Radar can tell you if your site has been cited in Perplexity’s answers. Competitive comparisons: Brand Radar also offers competitive share of voice – you can see, for example, how often your brand vs. a competitor’s brand appears in AI answers for your industry. It even helps discover “hidden competitors” – domains that frequently get cited by AI on topics of interest, which you might not have considered as SERP competitors ( ^[19] ) ( ^[20] ). A key metric here is the “AI impressions” or “AI mention count” – essentially, how many times your brand showed up in the AI results sample. Ahrefs has introduced concepts like Competitive AI Share (what portion of AI mentions in your niche belong to you vs competitors) ( ^[21] ) ( ^[19] ). For example, if in the topic “running shoes” AI answers cited Nike 40% of the time and Adidas 30% and your brand 5%, that quantifies your share of voice in AI for that topic. Brand Radar’s interface also tracks trends over time , so you can see if your AI visibility is rising or falling month to month. It even includes a “web visibility” index (traditional organic mentions on the web) and a “search demand” index (how search volume for your brand is trending) to correlate with your AI presence ( ^[22] ) ( ^[23] ). All of this augments the data from Google Search Console by focusing on where in AI ecosystems your brand appears, not just how many clicks you got. Semrush AI Toolkit: Semrush, another leading SEO platform, launched an AI Toolkit that automatically analyzes your brand’s presence across multiple AI platforms ( ^[24] ). It scans conversation data from ChatGPT, Google’s AI search (Gemini/AI Mode), Perplexity, and even niche platforms to determine: How visible is your brand in AI-driven search results? (It provides a “Market Share” percentage of AI search visibility ( ^[25] ).) How that visibility breaks down by platform – e.g., maybe you have strong presence in ChatGPT answers but weak in Google’s AI, or vice versa ( ^[10] ). Sentiment and context of mentions (more on that in the next section). The types of queries that lead to your brand being mentioned (informational, comparison, etc.). For example, Semrush’s toolkit can show that “Brand X appears in 20% of AI-generated answers about [category] on ChatGPT, 10% on Google, and 5% on Perplexity.” It might visualize this as a share-of-voice chart by platform ( ^[26] ). In one case study, the tool revealed that Warby Parker (an eyewear retailer) had far stronger visibility on ChatGPT than on Google’s SGE – implying their content was tuned well for conversational AI but needed improvement for Google’s AI summaries ( ^[10] ). This kind of insight is invaluable for prioritizing optimization efforts (you might need different tactics to win visibility on different AI engines). Additionally, Semrush’s AI Toolkit incorporates query intent analysis for AI search. It can categorize the intents behind questions mentioning your brand (e.g. “research vs. purchase intent”) ( ^[27] ) and help you understand what AI-driven consumers are looking for. It also offers competitive benchmarking – comparing your share of AI mentions and the sentiment of those mentions against competitors across platforms ( ^[28] ). The toolkit is essentially bringing classic SEO analytics (market share, sentiment, competitive analysis) into the AI answer space ( ^[29] ). Beyond these, a number of specialized startups have appeared with a pure focus on LLM visibility: Profound (TryProfound.com) – Focuses on analyzing how AI models talk about your brand. It claims to have sophisticated tech for sentiment analysis and even detecting AI hallucinations about your brand ( ^[30] ). Profound and similar tools use methods like synthetic queries (pre-programmed questions) to probe AI models and see what answers come up, then aggregate that data. Daydream (withdaydream.com) – An agency tool that, in partnership with a platform called Scrunch AI, monitors client brands in AI search results ( ^[31] ) ( ^[32] ). Daydream emphasizes tracking by prompt and persona – meaning they can see how a brand fares for different question phrasing and even different user personas. They provide analytics on citation frequency and positioning within answers ( ^[33] ). For instance, does the AI mention your brand as the top recommendation or just a footnote? Daydream’s blog notes that they track things like “how your brand performs by prompt, topic, and persona” and “citation tracking & source analysis” as part of their service ( ^[32] ). Good[ie] (HiGoodie.com) – Another entrant that reportedly helps manage model outputs about your brand. It can analyze prompts and help optimize how an AI might respond about you ( ^[34] ). All these tools share a common goal: make the invisible visible . They take the black-box outputs of AI systems and turn them into measurable data points for marketers – how often your brand appears, where it appears, what words surround it, and how that compares to competitors. This is analogous to how early SEO tools made Google’s opaque rankings trackable. Crucially, these AI visibility trackers should be used alongside traditional tools . Google Search Console and Analytics still tell you about traffic and on-site behavior. By combining them, you get a fuller story: GSC/Analytics: Are AI-driven changes affecting my traffic? (e.g. a drop in clicks for certain queries might coincide with the rollout of AI summaries.) AI visibility tools: Are we mentioned even when not clicked? (e.g. maybe clicks fell, but Brand Radar shows you were actually cited in the AI overview. Fewer clicks might simply mean the AI answered the question with your info without sending the user to you – a bittersweet outcome.) Together: If you see, for example, that you’re cited a lot by AI but getting few clicks, you may adjust your KPIs to value that brand exposure or find ways within the AI context to prompt clicks (perhaps by having compelling content that encourages the user to “learn more”). On the other hand, if you’re not appearing in AI answers at all and also losing traffic, that’s a sign you need to ramp up GEO efforts. It’s also important to track which AI platforms matter most for your audience. Recent data indicates that three AI platforms currently drive the bulk of referral traffic : ChatGPT, Google’s AI (Gemini/SGE), and Perplexity AI account for 98% of AI-sourced traffic to websites ( ^[35] ) ( ^[36] ). ChatGPT alone was the origin of about 50% of detectable AI referrals in early 2025, with Perplexity around ~30% and Google’s SGE (Gemini) about ~18% ( ^[37] ). This “big three” suggests that focusing on those platforms will cover most scenarios ( ^[36] ). However, each platform has a different style: e.g. ChatGPT (especially with browsing or plugins enabled) can directly cite and even click links, Perplexity always cites sources and tends to favor research-oriented content, while Google’s AI Overviews draw from high-authority web content and often favor well-established sources ( ^[38] ) ( ^[39] ). Bing’s AI (Bing Chat) is another notable platform, effectively an OpenAI GPT-4 variant that cites sources; while not broken out in the above stats, many SEO experts consider Bing Chat in the same category – and any tool that tracks “ChatGPT” answers may capture similar info for Bing, since Bing’s output is also reference-rich. Marketers should also keep an eye on emerging AI search entrants . For example, Anthropic’s Claude has been mentioned as being built into products like Apple’s Safari for AI Q&A ( ^[40] ) – meaning in the near future, a user asking their iPhone a question might get a Claude-powered answer with citations. If you operate in markets like China, Baidu’s ERNIE Bot integrates generative answers into search results, which introduces a whole parallel of GEO for Baidu. The tools for tracking those non-English AI mentions are nascent (and often internal to those markets), but the concept remains: you need to monitor your brand’s presence in whichever AI systems your customers use. In Europe or other regions, local players or open-source LLM-based search tools might gain traction, and keeping tabs on those (even if via manual testing) could give an edge. In summary, AI visibility tracking is now an essential complement to traditional SEO monitoring. You’ll want to answer questions like: How often is my site/brand being cited by AI, and is that trending up or down? Which AI platforms mention me the most (or least)? What queries or topics tend to include my brand in AI answers? (This can inform content strategy – you might discover AI loves citing one of your blog posts, suggesting that format or topic resonates with LLMs.) Who are my “AI competitors”? (These might be different from your normal SEO competitors. For instance, you might usually compete with Site A on Google, but find that in AI answers a government FAQ or a niche forum is what’s getting cited instead.) Is my overall share of voice in AI answers improving as I implement GEO tactics? (For example, after adding structured data or new FAQs, does Brand Radar show more mentions of you in the following months?) By combining data from GSC (for clicks/impressions) with AI mention data from new tools , you get a fuller visibility picture. An AI mention without a click still has value; a click without an AI mention means you came via traditional route. The ultimate goal is to maximize both – ensuring you are both mentioned by the AI and enticing the user to click through when appropriate. But even when the AI provides a complete answer, being referenced keeps you in the game (and maybe in the user’s consideration set for later).

13.3 Brand Sentiment and Context in AI Outputs

It’s not just if you’re mentioned by AI that matters – it’s also how you’re mentioned. In human conversations, context and tone shape perception, and the same is true for AI-generated content. Are AI summaries presenting your brand accurately and favorably? Measuring brand sentiment and context in AI outputs is a new frontier of reputation management. Marketers need to ensure that when AI discusses their brand, it’s painting the right picture and not inadvertently damaging the brand with misinformation or negative framing. Sentiment Analysis in AI Mentions: Leading SEO platforms have begun incorporating sentiment analysis specifically for AI-generated references. For example, the Semrush AI Toolkit evaluates whether mentions of your brand in AI answers are positive, neutral, or negative ( ^[41] ). It can give an overall sentiment score for your brand based on a sample of AI Q&A, and even identify the key drivers of sentiment ( ^[42] ). In one case, Semrush reported that Warby Parker had about 88% positive sentiment in AI mentions – the AI often described the brand in a favorable light, citing its strengths like affordable prices and home try-on program ( ^[43] ). The negative mentions (the remaining ~12%) touched on things like limited in-store selection ( ^[43] ). This kind of analysis is incredibly useful: it tells you what aspects of your brand the AI is emphasizing. Are the AIs highlighting your selling points or zeroing in on an old criticism? If an AI consistently describes your product as “expensive” or “difficult to use,” that’s clearly a problem – it might deter users before they even reach your site. Conversely, if AI assistants are extolling your key benefits (perhaps from reviews or content they’ve been trained on), that’s valuable positive exposure. Tracking sentiment allows you to quantify this. Some GEO monitoring tools create dashboards that break down the share of positive vs negative descriptors used alongside your brand ( ^[41] ). Marketers should monitor shifts in this sentiment over time or differences between platforms. It’s possible, for instance, that ChatGPT’s older training data might mention some past issue with your product that has since been resolved, whereas a real-time system like Perplexity (which fetches current info) might reflect newer, more positive reviews. Knowing that discrepancy can inform your content strategy (e.g., do more PR to push fresh positive content that AI will pick up) or even direct engagement (perhaps issuing clarifications via content if an AI is propagating a misconception). Accuracy and Hallucinations: Hand-in-hand with sentiment is accuracy . AI models can “hallucinate” – in other words, fabricate or err in the details. This can be harmless (like slightly misquoting a stat) or it can be seriously damaging (like incorrectly stating a fact about your company). There have been early incidents of AI chatbots providing false information about individuals or businesses. For example, a chatbot might mix up two companies with similar names or might regurgitate an outdated piece of information (such as “Brand X’s CEO resigned amid scandal” when in reality that was a different company or long in the past). Monitoring tools can help flag such occurrences. Some specialized platforms (like Profound or others) claim to identify hallucinations about your brand ( ^[44] ) – essentially catching when the AI says something that isn’t supported by known data. While it’s tricky to automate fully, one can at least periodically audit AI outputs for factual accuracy regarding your brand. If you discover common errors (for instance, the AI often gets your pricing or product specs wrong), you may need to update your content or FAQs to clarify those points or even use techniques like structured data to feed correct info. A proactive approach some companies use is fine-tuning or prompt-testing models for brand-specific Q&A . As noted in the Andreessen Horowitz analysis, new platforms for GEO are fine-tuning their own versions of models specifically to see how brands are portrayed ( ^[45] ). They simulate typical user prompts about the brand (e.g., “Is [Brand] reliable?”, “What does [Brand] do?”, “Compare [Brand] vs [Competitor]”), and then they analyze the output for tone and accuracy. The outputs can then be scored for sentiment or checked against fact. If the AI’s answer says something like “[Brand] is one of the more expensive options and has limited customer support hours” , a brand might recognize both a perception issue (expensive) and possibly an accuracy issue (maybe their support hours aren’t limited anymore). Some tools even allow you to input a desired brand positioning and see how close or far the AI’s answers are from that ideal. For example, if you want to be known as “innovative and affordable,” you’d hope AI descriptions use those or similar words. If not, it suggests your content and PR might not be emphasizing the right messaging, or that negative content is outweighing positive in the model’s training mix. “Model perception” is now something to cultivate: one SEO strategist phrased it as ensuring “the model remembers you in the right way.” In other words, how your brand is encoded in the AI’s knowledge is the new battleground ( ^[46] ). Context and Framing: Beyond sentiment polarity, context matters. Are you being mentioned as a top recommendation or as an also-ran? Is the AI summarizing your content accurately or taking it out of context? These qualitative aspects require looking at the actual AI answers. For instance, let’s say a user asks, “What are the downsides of Product A?” If your company makes Product A, you’d be very interested in what the AI says. It might list some cons that are exaggerated or outdated. Through prompt-testing, you discover the AI answer references an old review or a competitor’s blog that paints your product in a harsh light. That’s a cue to create fresh content addressing those points or to seek more positive coverage to outweigh the negative. Conversely, if the AI is asked, “Which product is better, [Yours] or [Competitor]?” – how does it respond? Many AIs will try to give a balanced view, but the details it chooses to mention can sway the reader. Monitoring these comparative answers is crucial for marketers. If the AI consistently highlights your strengths (e.g. “[Your Product] has a more intuitive interface and lower cost” ), great. If it highlights weaknesses, you know what you need to improve or clarify in your materials. A real-world example: the A16Z GEO report shared a case where Canada Goose (an apparel brand) analyzed how often and in what context LLMs referenced it ( ^[47] ). The interesting find was that beyond just product features, the brand wanted to know if AIs would “spontaneously mention” Canada Goose when talking about winter jackets in general – an indicator of brand prominence ( ^[47] ). For them, it wasn’t only about sentiment but about whether the brand name surfaces at all without being directly prompted . This is another angle of context: unaided brand recall by AI . If people ask generally for recommendations (“best winter coat brands”), does the AI include you? If not, you have a brand awareness issue in the AI’s “mind.” If yes, is it associating the right qualities with you (e.g. durability, luxury, etc.)? Tools for sentiment & context: As mentioned, many of the AI tracking tools incorporate some sentiment analysis. Semrush’s toolkit, for instance, not only labels mentions as positive/neutral/negative but also identifies key sentiment drivers – essentially the topics or features that lead to positive or negative talk ( ^[42] ). In the Warby Parker example, the home try-on program and affordability drove positive sentiment, whereas limited selection and weak in-store presence were negatives ( ^[43] ). If such a report were about your brand, that’s gold: it’s telling you what messaging to amplify and what objections to mitigate in the eyes of AI (which often mirror common consumer perceptions). Other tools like the startup Goodvie/Goodie have pitched themselves as doing “model sentiment and prompt analysis” ( ^[34] ). This implies they might let you input a series of prompts and then they analyze not just if you’re mentioned but also the tone (perhaps by scoring the adjectives used, etc.). Similarly, Daydream’s integration with Scrunch AI boasts sentiment tracking so you can “discover how AI perceives your brand and track shifts over time.” ( ^[48] ). The ability to track shifts over time is important. Perhaps you launch a new product or a crisis hits your brand – how quickly and in what way does AI content reflect that? If sentiment plunges after a wave of bad news, you’ll see it in AI answers too. Or if you invest in thought leadership content and expert articles, you might see over months that AI answers become more favorable (because those positive signals have been ingested into the model’s knowledge). Watching these trend lines can validate the impact of your GEO and PR efforts. Actioning the insights: Measuring sentiment and context in AI outputs isn’t just for curiosity – it should feed back into strategy: If inaccuracies are found, correct them at the source . That might mean updating your website (AI like Bing and Perplexity might retrieve updated info), issuing press releases or FAQs, or even directly providing feedback to AI providers if possible. For instance, OpenAI has a feedback system for users to flag incorrect info; enough flags might influence retraining. Google’s SGE might slowly improve if the underlying web content becomes more accurate and consistent. If sentiment is negative due to genuine issues (e.g., many reviews complain about one aspect), that’s a business insight to fix the product or customer experience. If sentiment is negative due to outdated or skewed info, that’s an SEO/PR task: produce content to set the record straight. If sentiment is positive – promote it! Ensure those positive points (that AI is picking up) are reinforced on your site and marketing materials. Also, you might leverage the fact that AI likes certain content from your site to double down on that content strategy. If your brand isn’t being mentioned when it should (e.g., AI often mentions competitors in answers but not you), that indicates you might need to build more authority . This could mean creating high-quality content that others cite (so the AI sees your name associated with the topic), improving your site’s prominence (so the AI’s web crawl finds and trusts your content), or engaging in more traditional brand-building (since some models might be biased toward well-known names ( ^[38] )). In fact, an Ahrefs analysis in mid-2025 found that Google’s AI Overview tends to favor brands that have a lot of web mentions (correlation was high between web mention count and appearance in AI results), whereas ChatGPT and Perplexity were less brand-biased ( ^[49] ) ( ^[50] ). Google’s AI seems to lean on brand authority as a trust signal (much like its search did) ( ^[39] ). This means if you’re a lesser-known brand competing with giants, you may find Google’s AI less likely to mention you until you gain more overall prominence (backlinks, mentions, etc.), whereas other AI tools might be more “egalitarian” in citing niche sources. This insight again ties back to sentiment and context: Google might mention you primarily in contexts where you’re already recognized as a top authority (so your strategy might be to become a known authority through digital PR, etc.), whereas for a platform like Perplexity, even a single excellent article from you could get cited for a specific query. To wrap up, measuring sentiment and context in AI outputs is about safeguarding and shaping your brand’s narrative in this new medium. You should strive not only for a high reference rate, but also for a positive reference quality : The tone should align with your brand image (e.g., AI calls you “reliable and innovative” rather than “cheap” or “basic” unless those are intentional brand positions). The facts should be correct (no lingering outdated info). The associations should be favorable (e.g., you’re mentioned alongside other reputable brands, or for the product features you excel in). Through regular monitoring – using tools and manual spot-checks – you can catch issues early. In the same way companies have set up Google Alerts for their brand name on the web, they are now setting up AI alerts : essentially tracking when an AI mentions them and analyzing the output. By doing so, you maintain control over your brand's representation in AI-driven channels and can adjust your GEO tactics to keep that representation positive.

13.4 Conversion and Engagement Shifts in the AI Era

The rise of AI-generated answers is not only altering how we get visibility; it’s also changing the nature of the traffic that does reach our websites. Marketers may notice shifts in traditional metrics like total organic traffic, conversion rates, on-site engagement, and lead quality as AI becomes the first touchpoint. In this section, we explore how GEO impacts user behavior and downstream conversions – and why quality may trump quantity going forward. Traffic Volume vs. Intent: One immediate effect of AI answers is a potential decline in raw organic clicks for certain queries. If a user’s question is fully answered in an AI snippet, they might not feel the need to click any result. Early data from Search Engine Land and others have predicted significant drops in organic clicks (some estimates ranged from 18% up to 64% for certain queries) due to SGE-style answers occupying the top of the SERP ( ^[51] ) ( ^[52] ). Indeed, an internal Google study noted that when the AI snapshot is present, the classic results get pushed far down, often below the fold on mobile ( ^[53] ). However, focusing only on the loss of clicks can be misleading. It is crucial to ask: which clicks are being lost? Typically, it’s the quick fact-finding or superficial queries – the kind where the user just needed a definition, a simple piece of advice, or a list of options. When those low-intent, informational needs are satisfied by AI, the clicks that remain are often the higher-intent visits : users who have follow-up questions, want to dive deeper, or evaluate options in detail. In other words, the AI may filter out the top-of-funnel browsers, leaving you with more mid- or bottom-of-funnel visitors. For example, someone searching “How to unclog a drain” might get the step-by-step from an AI overview and never click, whereas someone searching “Plumber near me cost” or “Order BrandX drain snake” likely has a stronger intent and might click through to sites for specifics. Marketers are observing that while total visit counts from organic search could plateau or drop, the engagement and conversion rates of those who do click can rise . One digital agency analysis suggested that as SGE rolls out, “this could lead to increased conversion rates and a higher quality of organic search traffic.” ( ^[54] ). Why? Because the people who make it past the AI answers to your site are self-selecting as more motivated. They either weren’t fully satisfied with the summary or are far enough along in their journey that they need the detail/transaction your site offers. In practice, you might see metrics like: Longer average session duration or lower bounce rate from organic users, because those who click genuinely want more information or are closer to decision-making. Higher conversion rate (be it form fills, product purchases, etc.) from organic, as casual information-seekers have been siphoned off by the AI. Your site visits might be fewer but “meatier.” A concrete example comes from anecdotal evidence in the tech sector: Vercel, a developer platform, found that around 10% of their new sign-ups were coming through ChatGPT’s referrals ( ^[55] ). This implies that users who learned about Vercel via ChatGPT (likely through code suggestions or answers mentioning Vercel) were clicking through and converting at a meaningful rate. Those referred users presumably already had high intent – ChatGPT might have recommended Vercel as a solution, so by the time the user hit the site, they were primed to sign up. This kind of referral might be low in volume but high in conversion efficiency. Another study by Ahrefs noted a subtle but important trend: on average, keywords that trigger AI Overviews had higher zero-click rates (users not clicking anything) but when looking at the same keywords before vs after AI was introduced, the presence of AI **slightly decreased the zero-click rate** ( ^[56] ) ( ^[57] ). In other words, some users who previously might not have clicked anything at all were now clicking on an AI-cited source. This suggests that AI answers can generate curiosity or credibility that draws certain users to click through for more (especially if the AI snippet includes an intriguing tidbit or partial info). So, while AI overviews reduce the need for many users to click, they also inspire a subset of users to click specific sources. The net effect on clicks is complex, but from a conversion standpoint, those who do click may be more informed (they’ve read a summary already) and thus further down the funnel. Shifting KPIs: Given these dynamics, companies are beginning to adjust their KPIs and success metrics for organic search in the AI era: Instead of obsessing purely over total organic sessions , more attention is paid to conversion-related metrics : lead volume from organic, revenue from organic, ROI per visit, etc. If you lose 20% of your organic traffic but your leads or sales remain the same, it means the traffic you lost was largely informational window-shoppers. Your marketing success might still be intact (or even improved in efficiency). Brand visibility metrics (like reference rate, as discussed) become a parallel KPI. Even if traffic drops, you can show stakeholders that “We are being mentioned in 30% of AI answers about [product category] – that’s brand exposure we wouldn’t have otherwise.” You might track how that correlates with branded searches or direct traffic. For example, maybe your direct visits or brand-name searches increase because people saw you mentioned by AI and later come directly to you. These indirect conversions might not be immediately obvious, but they are part of GEO success. (One hint of this was mentioned in the Ahrefs traffic study: they noted that AI conversions may be underestimated because a user could be influenced by an AI answer and convert later without a traceable click ( ^[58] ). For instance, an AI might recommend a specific product, and the user goes straight to Amazon or the brand’s site later to buy it – no referral tag, but the AI answer started the chain.) On-site behavior metrics : Are pages seeing longer dwell time? Is the scroll depth deeper? These can signal that the nature of organic visits is now more serious. If someone already got the basics from AI, when they come to you they likely want the nitty-gritty – and indeed might spend more time consuming that content or evaluating products. All this is to say, marketers should pivot from a pure volume mindset to a quality mindset for organic reach . High-level executives might initially panic seeing a drop in Google traffic due to AI. It’s our job to contextualize that with quality metrics. For example: “Yes, traffic on these how-to articles fell after AI answers started showing, but the traffic we’re still getting is 2× more likely to convert. Also, our brand is still present in those AI answers as a cited authority, which has intangible value and likely drives some direct visits.” Engagement and Funnel Position: When basic questions are answered by AI, users coming in might be later in the funnel: Someone who asks an AI, “What are the top 5 project management tools?” gets a quick list (with perhaps your brand included). The user then asks follow-up, “Which of those is best for a small team?” The AI elaborates, maybe citing some comparison from your blog. By the time they click your site, they might be at the consideration/evaluation stage rather than the discovery stage. So, when they land on your page, they might immediately seek free trial info or pricing – because they already know who you are and some pros/cons, courtesy of the AI introduction. Alternatively, they might have only heard of you through the AI’s answer. This means your site needs to capture them quickly with what you want them to do (this is where good UX and on-page prompts matter more than ever). If AI delivers a user to you, that user has a certain expectation set by the question they asked. Ensure the landing content aligns with that intent (this hasn’t changed from SEO best practices, but the questions might be more specific now). Marketers should monitor if the landing page mix for organic traffic changes. Are fewer people entering on your generic blog posts and more entering on product pages or deep comparison pages? If so, that suggests AI handled the generic stuff, and users click when they’re ready for specifics. You might see organic homepage entrances drop (because AI answered “who is Company X” so they didn’t click your homepage) but product page entrances rise (they clicked on a product link from the AI because they wanted details or to buy). Keeping an eye on this can validate the idea that the remaining clicks are more intent-rich . Conversion Rate Optimization (CRO) Adjustments: With potentially more qualified traffic, you might achieve better conversion rates – but don’t leave it to chance. Double down on CRO for organic entry pages. If visitors now are later-stage, ensure your calls-to-action meet them. For example, if a significant chunk of organic visitors are coming after an AI recommendation, they might be primed to sign up – so make that process smooth and obvious on the landing page. On informational pages that still get traffic, consider that those visitors likely have very specific questions that maybe the AI couldn't fully satisfy (or else they wouldn’t be there). It may help to include clear FAQ sections, or interactive elements, to serve those advanced needs. Also, consider measuring micro-conversions that reflect engagement, not just macro conversions (sales/leads). Micro-conversions like newsletter sign-ups, time spent on site, number of pages viewed, or even interactive elements used (like a product compare tool, calculator, etc.) can show that even if absolute users are fewer, they’re engaging more deeply. These are good signs that your content is attracting a more invested audience. Adjusting marketing mix and expectations: If you do find that AI is cutting a big chunk of your trivial-queries traffic, you might reallocate some effort: Perhaps invest less in writing basic how-to articles that an AI will likely summarize, and more in very original, expert or data-rich content that AI might cite or that users will seek out for depth. (This aligns with the broader strategy of creating content AI can’t easily generate on its own – discussed in Chapter 9 on content strategy.) Focus on branding . If fewer people click to discover you, you want to ensure your name is recognizable and trusted when the AI mentions it. Off-page GEO (Chapter 11) becomes relevant: building authority through PR, communities, etc., so that when an AI chooses sources, it favors or at least includes you. This indirectly helps conversion because users are likelier to click known brands in the AI citations. A study by Nielsen decades ago showed brand familiarity boosts click-through and conversion, and a similar effect likely happens in AI results – if the user sees a source they know (or have seen referenced often), they might gravitate to it. This is speculation but supported by Google’s bias toward brands ( ^[39] ); big brands get referenced and likely get the clicks that do happen. So smaller players need to work harder to build that credibility. Case in point – quality over quantity: Let’s illustrate with a hypothetical scenario: Before AI, your site got 1,000 organic visits/month, converting 20 leads (2% conversion). After AI overviews roll out, you drop to 700 visits/month. At first, this looks bad. But upon analysis, those 700 visits convert 21 leads (3% conversion). Your total leads even ticked up slightly, and conversion rate improved markedly. This suggests the AI filtered out 300 visits that were unlikely to convert anyway. Additionally, you find that your brand was cited in 100 AI answers that month – exposures that might not have immediate clicks but contribute to brand awareness. So while old-school metrics would show a 30% traffic drop, the smarter interpretation is that you maintained or even improved performance with fewer, more qualified visitors while also gaining “mindshare impressions” via AI. Explaining this to stakeholders is crucial so that they understand GEO success isn’t about raw traffic growth in the same way SEO was, but about maintaining outcomes (sales, leads) and presence in a changing search landscape . Another forward-looking metric is engagement shift : if AI handles simple queries, perhaps the traffic that ends up on your site is looking for interactive or value-added experiences (things an AI can’t do well, like provide tools, live data, personal consultations, etc.). Measuring usage of those features (like how many use your online calculator or start a live chat) can show that your site is capturing the users who need more than a static answer. ROI and Attribution Considerations: With AI playing middleman, attribution might get muddy. A user could be influenced by an AI recommendation and later come through organic search by typing your brand (making it look like a direct or branded search conversion, when the assist was AI). To the extent possible, gather anecdotal feedback: ask your leads or customers how they heard about you. You might start hearing “I was chatting with ChatGPT and it suggested your tool,” or “I saw your product mentioned in the new Bing.” This qualitative insight can reinforce that AI references are driving real business (just indirectly). Some companies have even started including an “How did you hear about us?” option for “AI assistant (e.g., ChatGPT, Bing AI, etc.)” in forms to track this emerging channel. In summary, don’t panic if you see fewer clicks – instead, measure what those clicks do, and measure how your brand is surfacing in AI-driven pathways that don’t result in a click. You may need to adjust your success benchmarks : for instance, instead of saying “we aim to grow organic traffic 10%,” you might say “we aim to maintain organic lead volume, while also achieving X% presence in top AI answers for our category.” Internally, this is a big mindset shift, but it’s one that forefront marketers are adopting. They understand that in the AI era, being the answer is just as valuable as getting the click , and the traffic that does come through is more valuable than ever.

13.5 Continuous Testing and Experimentation: Adapting to the GEO Frontier

Generative AI in search is still a fast-evolving field. What gets your content referenced today might not work tomorrow after the next model update. Likewise, new AI features and platforms are emerging rapidly. Therefore, a culture of continuous testing and learning is vital for GEO success. Marketers should establish processes to regularly experiment, monitor, and refine their GEO strategies – in essence, to treat GEO optimization as an ongoing cycle, much like traditional SEO where you continuously tweak and observe rankings. Here we discuss how to set up that test-and-learn framework for GEO. Experimenting with Content Changes: One practical approach is to make targeted content optimizations and then see if AI responses reflect the change . For example, suppose you notice that Google’s AI overview for a query about your industry never cites your site. You decide to update one of your relevant pages with a new section (perhaps a concise answer to that query, with clear phrasing and maybe schema markup). Once that page is reindexed, you watch to see if the AI overview now starts pulling from it. This can be done by manually triggering the AI overview (if you have SGE access) or by checking a tool like Brand Radar for that query in the next data refresh. If you succeed and your site becomes one of the cited sources, that’s a win – and a learning that the change made a difference. If not, that’s also instructive: maybe the content needs further improvement or a different approach. Similarly, you might add an FAQ section to key pages, anticipating common questions users ask AI, and then later ask those questions to ChatGPT or Bing to see if your new FAQ content gets used in the answer. If you see phrases from your FAQ mirrored in the AI’s answer, bingo – your content is influencing the model output. SEO professionals have started doing this kind of testing routinely: essentially, feeding likely prompts to various AI systems after each significant site change. This is analogous to how one might do A/B testing on webpages; here you’re testing how the “AI audience” responds to content tweaks. Synthetic Prompt Testing: Some GEO agencies formalize this through synthetic prompt testing frameworks . They will create a list of representative user queries (prompts) relevant to the business – covering various funnel stages and question phrasings – and then programmatically or manually run those prompts on a schedule across different AI engines (ChatGPT, Bing Chat, Bard/Gemini, Claude, Perplexity, etc.). The outputs are then parsed to check for brand mentions, citations, sentiment, etc. Monstrous Media Group, for instance, describes that they use test queries to evaluate whether a client’s content is being referenced in AI responses, and if not, they adapt the content ( ^[59] ). This iterative testing approach allows you to measure improvement: e.g., last quarter, our brand showed up in 2/20 test prompts on ChatGPT; after optimizing content and building some links, it shows up in 5/20 prompts. It’s a controlled way to gauge progress in a world where we can’t directly see “rankings.” There are even tools emerging to automate prompt testing and result analysis (some SEO suites might add “Prompt Rank Tracker” features in the future). But you don’t need to wait for fancy tools – you can DIY by periodically querying AI with key questions and logging the results. Many SEO teams now have a standing monthly or bi-weekly task: “Check our presence on AI answers for top use-case questions.” This could be as simple as using ChatGPT’s browsing tool or Bing’s chat and seeing if your site/article is mentioned or linked. Model Updates = New Algorithm Updates: Keep in mind that AI models update (or are replaced/upgraded) periodically, akin to Google’s algorithm updates. OpenAI might release a new GPT version, Google will update its Gemini model, etc. Each update could change how content is interpreted or which sources are favored. For example, maybe an update significantly improves citing recent information, so suddenly sites with the freshest content gain an edge in AI answers. Or an update fixes a prior hallucination issue, altering how your brand is described. Because of this, you should: Monitor industry news on AI model updates (OpenAI, Google, Anthropic, etc. often announce major changes). After a known update, re-run your prompt tests. See if your reference rate or answer patterns changed. It’s reminiscent of how an SEO might check their rankings after a Google core update – here you check your AI mentions after a model update. Have a plan to quickly respond. If an update wipes out many of your mentions (maybe the AI is now citing Wikipedia for everything, hypothetically), you may need to adjust strategy (perhaps focus on being the source that Wikipedia or other high-authority sites cite, or provide more unique data that stands out). An insightful quote from the a16z GEO report: “With every major model update, we risk relearning (or unlearning) how to best interact with these systems… LLM providers are still tuning the rules behind what their models cite.” ( ^[60] ). This underlines the importance of staying agile and not assuming today’s tactics will always work. Just as SEO had its Panda, Penguin, and countless other updates that required course-corrections, the GEO world will have its own twists and turns. Cross-Platform Testing: Don’t limit testing to one AI platform. Each has its quirks: Test Google’s AI (SGE/Gemini) for how it constructs overviews with your content. Are there queries where it could cite you but isn’t? Why might that be (is your content not as crawlable or is another site just more directly answering the question)? Test Bing Chat – it might use a different style. Does it find your content? Bing often provides up to 3–4 citations per response; see if you’re among them for relevant questions. Test voice assistants or multimodal AI if relevant. For instance, if you care about Alexa or Siri (which might integrate more generative capabilities), how do they respond about your brand? This might not be as open to custom prompts yet, but keep an eye on it as these assistants evolve. Test international or domain-specific AIs if you have markets there. If you operate in, say, Japan, you might test Naver’s HyperCLOVA or other local AI tools with relevant prompts (though accessing some might be tricky without native accounts). The idea is to not be blindsided by an AI that’s popular in a region you serve. Internal Process and Collaboration: Implementing continuous GEO testing can be made part of your ongoing SEO or content workflow: Schedule regular reviews: e.g., a monthly “AI Visibility Report” that your team generates. It could include data from Brand Radar/Semrush on brand mentions, plus qualitative observations from manual tests. This keeps GEO on the radar (pun intended) in the same way monthly SEO reports do. Interdisciplinary input: GEO straddles SEO, content, PR, and even product. Consider having your customer support or sales team glance at AI outputs too – they might catch something like the AI giving outdated pricing or feature info. That insight can trigger a content update or a clarification on your site. Brainstorm experiments: Encourage the team to hypothesize and test. For example, “We aren’t being mentioned for [topic]. What if we publish a unique research study on that topic – will the AI pick it up?” Then do it and watch. Or, “Our competitor is always cited for definition of X. Let’s create a more comprehensive definition and see if we can replace them.” Treat it like scientists running lab experiments. One particular tactic is embedding likely user prompts into your content (which might have been discussed in Chapter 12 on prompt optimization). This means literally taking common questions users ask AI and including them (and their answers) on your site – in a FAQ section, in headings, etc. By doing so, you increase the chance the AI will use your text when that question is asked. After embedding, you test again: ask the AI that question, see if your phrasing appears. This is a direct way to validate the impact of prompt-oriented content optimization. Learning from Failures and Successes: If you test something and it doesn’t work – that’s still a learning. GEO is new for everyone, so sharing lessons (even internally) is valuable. Maybe you find that adding a certain schema markup got your content picked up by the AI overview (there have been theories that schema like HowTo or FAQ might influence SGE). Or you find that the AI ignored your page until it gained a few external links, at which point it started citing it – suggesting authority threshold matters. These insights help refine your approach. Also, keep an ear to the ground: SEO forums, case studies (like the Diggity Marketing case study where an agency got a 2,300% increase in AI-driven traffic ( ^[61] )), and industry webinars are now popping up specifically about GEO experiments. For example, some practitioners share how they got a featured snippet and an AI mention by structuring content in a certain way. Being part of that conversation and knowledge-sharing will accelerate your success. The landscape in 2024/2025 is such that everyone is experimenting , so there’s a lot of community learning to tap into (and contribute to). Governance and Ethical Testing: In your rush to optimize for AI, maintain ethical standards. Avoid trying to trick the AI with spammy tactics (like stuffing a ton of keywords in hidden text hoping the AI training picks it up – that could backfire or be considered manipulation). Chapter 12 likely covered avoiding black-hat tactics; it applies here in testing too. If you use a tool to mass-query an AI, be mindful of the platform’s usage policies. Some AI platforms might restrict automated querying. Use official APIs or approved methods if you want to scale up prompt testing, to avoid being blocked. Closing the Loop: Finally, ensure there is a feedback loop from what you learn back into your strategy: If a test shows improvement, operationalize it (e.g., “We tested adding summaries at the top of articles; it got us cited more on Perplexity. Let’s roll this out across more content.”). If you discover new queries that people ask AIs (maybe via tools or from logs if available), feed that into content planning. Incorporate AI visibility as a goal in your content briefs. For instance, when creating a piece, define which AI queries you hope it will answer, and structure it accordingly (Q&A format, clear sourcing, etc.). Then after publish and indexing, test those queries to see if it worked. Just as early SEO had an experimental nature (“Let’s try this keyword density or that title tag and see what ranks”), early GEO demands creativity and iteration. By establishing a rigorous yet flexible testing regimen, you’ll keep your GEO efforts aligned with the moving target of AI algorithms. The companies that excel will be those that learn faster and adapt continuously – turning GEO from a one-time project into a sustained capability. In conclusion , measuring GEO success requires both old and new lenses : using some traditional analytics in new ways (focusing on conversion quality, using Search Console’s data with an AI filter in mind) and embracing brand-new tools and metrics (reference rate, AI share of voice, AI sentiment). It’s a more complex picture of performance, but a richer one. By tracking references, sentiment, and engagement shifts – and by relentlessly testing and learning – online marketing professionals can ensure that their brands not only survive but thrive in the age of AI-driven search. The next and final chapter will look ahead at future trends and how to build on these strategies for long-term success, but armed with the metrics and tools covered here in Chapter 13, you are well-equipped to quantify and optimize your GEO initiatives today.

References

^[1] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[2] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[3] Journal.Withdaydream.Com Article - Journal.Withdaydream.Com URL: https://journal.withdaydream.com/p/llm-optimization-for-all-daydream-clients

^[4] Monstrousmediagroup.Com Article - Monstrousmediagroup.Com URL: https://monstrousmediagroup.com/services/generative-engine-optimization-geo

^[5] www.kailos.ai - Kailos.Ai URL: https://www.kailos.ai/insights/beyond-seo-a-guide-to-generative-engine-optimization-geo

^[6] Monstrousmediagroup.Com Article - Monstrousmediagroup.Com URL: https://monstrousmediagroup.com/services/generative-engine-optimization-geo

^[7] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[8] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[9] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[10] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[11] Monstrousmediagroup.Com Article - Monstrousmediagroup.Com URL: https://monstrousmediagroup.com/services/generative-engine-optimization-geo

^[12] Monstrousmediagroup.Com Article - Monstrousmediagroup.Com URL: https://monstrousmediagroup.com/services/generative-engine-optimization-geo

^[13] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/google-ai-mode-traffic-data-search-console-457076

^[14] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/google-ai-mode-traffic-data-search-console-457076

^[15] Search Engine Land Article - Search Engine Land URL: https://searchengineland.com/google-ai-mode-traffic-data-search-console-457076

^[16] Ahrefs Article - Ahrefs URL: https://ahrefs.com/brand-radar

^[17] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[18] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[19] Ahrefs Article - Ahrefs URL: https://ahrefs.com/brand-radar

^[20] Ahrefs Article - Ahrefs URL: https://ahrefs.com/brand-radar

^[21] Ahrefs Article - Ahrefs URL: https://ahrefs.com/brand-radar

^[22] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[23] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/new-features-june-2025

^[24] www.semrush.com - Semrush.Com URL: https://www.semrush.com/kb/1493-ai-toolkit

^[25] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[26] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[27] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[28] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[29] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[30] Sourceforge.Net Article - Sourceforge.Net URL: https://sourceforge.net/software/product/AI-Brand-Tracking/alternatives

^[31] Journal.Withdaydream.Com Article - Journal.Withdaydream.Com URL: https://journal.withdaydream.com/p/llm-optimization-for-all-daydream-clients

^[32] Journal.Withdaydream.Com Article - Journal.Withdaydream.Com URL: https://journal.withdaydream.com/p/llm-optimization-for-all-daydream-clients

^[33] Journal.Withdaydream.Com Article - Journal.Withdaydream.Com URL: https://journal.withdaydream.com/p/llm-optimization-for-all-daydream-clients

^[34] www.kailos.ai - Kailos.Ai URL: https://www.kailos.ai/insights/beyond-seo-a-guide-to-generative-engine-optimization-geo

^[35] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[36] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[37] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[38] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search

^[39] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search

^[40] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[41] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[42] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[43] Avenuez.Com Article - Avenuez.Com URL: https://avenuez.com/blog/semrush-new-ai-toolkit-a-complete-review

^[44] Sourceforge.Net Article - Sourceforge.Net URL: https://sourceforge.net/software/product/AI-Brand-Tracking/alternatives

^[45] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[46] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[47] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[48] Journal.Withdaydream.Com Article - Journal.Withdaydream.Com URL: https://journal.withdaydream.com/p/llm-optimization-for-all-daydream-clients

^[49] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search

^[50] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search

^[51] Pilotdigital.Com Article - Pilotdigital.Com URL: https://pilotdigital.com/blog/google-generative-search-sge-and-its-effect-on-organic-traffic

^[52] www.warc.com - Warc.Com URL: https://www.warc.com/newsandopinion/opinion/did-google-just-steal-all-your-organic-traffic-introducing-the-search-generative-experience-sge/en-gb/6505

^[53] www.climbingtrees.com - Climbingtrees.Com URL: https://www.climbingtrees.com/google-sge-will-reshape-organic-traffic-growth

^[54] www.climbingtrees.com - Climbingtrees.Com URL: https://www.climbingtrees.com/google-sge-will-reshape-organic-traffic-growth

^[55] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[56] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/semrush-ai-overviews-study

^[57] www.semrush.com - Semrush.Com URL: https://www.semrush.com/blog/semrush-ai-overviews-study

^[58] Ahrefs Article - Ahrefs URL: https://ahrefs.com/blog/ai-traffic-study

^[59] Monstrousmediagroup.Com Article - Monstrousmediagroup.Com URL: https://monstrousmediagroup.com/services/generative-engine-optimization-geo

^[60] A16Z.Com Article - A16Z.Com URL: https://a16z.com/geo-over-seo

^[61] Diggitymarketing.Com Article - Diggitymarketing.Com URL: https://diggitymarketing.com/ai-overviews-seo-case-study

14. Future Trends – GEO and the Next Frontiers of Search

The landscape of search is poised to undergo transformative changes in the coming years, driven by advances in AI and shifting user behaviors. Generative Engine Optimization (GEO) will need to evolve in step with these trends. This chapter explores five key frontiers shaping the future of search and what they mean for online marketing professionals: personal AI assistants , multimodal search and answers , voice search 2.0 , evolving AI algorithms and transparency , and the enduring role of human creativity . Each section delves into emerging developments and provides guidance on how to adapt content and SEO strategy accordingly.

14.1 Personal AI Assistants: The Personalized Search Revolution

One of the most significant shifts on the horizon is the rise of personal AI assistants – AI agents tailored to individual users, drawing on personal data and preferences in addition to general web knowledge. These AI “co-pilots” are changing how people find information and make decisions, effectively becoming intermediaries between users and the web. In this future, consumers may rely less on traditional search engines and more on AI-driven agents like smart assistants, chatbots, or custom AI models that act on their behalf ( ^[1] ).

14.1.1 AI Agents in Everyday Life: Changing Search Behavior

AI assistants are already gaining traction in daily life. A 2024 McKinsey study found 41% of Gen Z consumers rely on AI-driven assistants for shopping and task management ( ^[1] ) – from getting recommendations on what to buy to managing calendars. This trend spans demographics and is expected to grow. These assistants (e.g. OpenAI’s ChatGPT, Google’s Bard, Amazon’s Alexa, Apple’s Siri, or newcomer apps like Inflection’s Pi) are increasingly capable of complex reasoning and personalization. They don’t just fetch information; they execute tasks and influence decisions autonomously ( ^[2] ). For example, instead of a user performing multiple Google searches to plan a trip, a personal AI agent could handle the entire process: considering the user’s past travel preferences, budget from personal finance apps, and live flight data to propose a tailored itinerary. In effect, the AI agent bypasses traditional search engines , drawing on content it has pre-ingested or fetched via APIs to provide answers or actions ( ^[3] ). This has profound implications: content that isn’t accessible or trusted by these AI agents will be invisible in this new channel of discovery. Real-world example: Microsoft’s Copilot in Windows 11 and Office 365 is an early example of a broad personal AI assistant for both consumers and enterprise workers. By late 2024, nearly 70% of Fortune 500 companies had adopted Microsoft 365 Copilot to assist employees with tasks like drafting emails, summarizing documents, or answering questions from internal data ( ^[4] ). These AI copilots combine a large language model (OpenAI GPT-4 in this case) with user-specific context (emails, files, calendar) to deliver highly relevant answers. Similarly, Google Assistant with Bard/Gemini integration is enabling more personalized help on mobile devices. Google’s Assistant can now use its Gemini AI model to understand images and context from your screen, and even draw on your personal Gmail or documents (if permissions allow) to answer queries ( ^[5] ) ( ^[6] ). For instance, you might ask on your phone: “ Remind me what my flight details are and what’s a good restaurant near my hotel? ” – a Gemini-powered Assistant could parse your email for the flight info and cross-reference Google Maps for restaurant reviews, all in one conversational reply. Such scenarios illustrate how personalized AI search is becoming: the AI taps both private data (user’s emails, calendar, past behavior) and public web data to craft an answer. From a marketer’s perspective, this means the traditional funnel of users searching on Google and clicking a link is being disrupted. The AI assistant might never display a list of blue links ; instead, it gives a spoken or written answer synthesized from multiple sources. As one AI SEO expert put it, by 2025 failing to be present in the “trusted dataset” of AI assistants will be akin to not ranking in Google today ( ^[7] ). GEO strategy, therefore, must consider “AI visibility” – ensuring your content is among the sources these AI agents draw upon.

14.1.2 Integration of Personal Data and Enterprise AI

– Personal AI assistants aren’t just consumer tools; businesses are adopting them as well. Enterprise AI assistants (like IBM Watson Assistant or custom chatbots fine-tuned on company data) are helping employees and customers get information more efficiently. Microsoft reports that generative AI adoption reached 75% of companies in 2024 , with many seeing significant ROI from AI assistants handling routine queries ( ^[8] ). For example, a large consulting firm used a custom AI agent to automate parts of client onboarding, cutting lead time by 90% ( ^[9] ). These enterprise agents often combine internal knowledge bases with web information. For marketers, this opens a new front: B2B GEO . If companies use AI assistants to find vendors or solutions (instead of manually searching), you need to ensure your thought leadership content, case studies, and product info are accessible to those AI systems. This could mean providing structured data feeds or APIs to enterprise clients, or simply continuing to produce high-authority content that an AI agent evaluating options would deem credible. Moreover, personal assistants in home and life – from smart home hubs (Alexa, Google Nest Hub) to car assistants (Tesla’s planned integration of xAI’s Grok chatbot into vehicles) – will draw on web content for answers. Amazon’s revamp of Alexa with generative AI is instructive: in late 2024 Amazon announced “ Alexa Next Gen ,” powered by Anthropic’s Claude model, to handle more complex questions ( ^[10] ) ( ^[11] ). This new Alexa can converse more naturally and provide detailed answers (it even has a paid tier for advanced AI capabilities) – essentially becoming a ChatGPT-like search agent you talk to. Amazon hopes this will finally get users to engage in shopping and recommendations via Alexa, something that earlier voice search hadn’t achieved ( ^[12] ). Early signals suggest voice commerce may indeed pick up: voice assistant users are 33% more likely to have made an online purchase in the past week than the average consumer ( ^[13] ) ( ^[14] ). As these assistants get better at transactions, optimizing your product listings for voice/AIs or integrating with assistant platforms (e.g. ensuring your retail site’s products are indexed in Amazon’s or Google’s shopping graphs that Alexa/Assistant pull from) will be critical.

14.1.3 Optimizing Content for Personal AI Assistants

– How do we adapt GEO strategies for this agent-dominated world? The fundamentals of Answer Engine Optimization (AEO) still apply but must be taken further. AI assistants crave content that is structured, concise, and semantically rich ( ^[15] ). Key tactics include: Schema Markup and Structured Data: Use structured metadata (JSON-LD, schema.org) to help AI parse your site’s content. Personal assistants often consult knowledge graphs and structured databases. Marking up product information, reviews, business details, FAQ answers, etc., increases the chance that an AI agent finds and trusts your data. For instance, a personal assistant booking travel might pull rates or availability from schema-tagged data on your site rather than scraping plain text. FAQ and Conversational Content: Anticipate the questions users might verbally ask their assistants. Implement FAQ sections with straightforward Q&A pairs. Well-crafted FAQ content is highly valued because AI can directly quote it as an answer ( ^[16] ). If a smart speaker is asked “ What’s the shelf life of this supplement? ”, a concise answer in your FAQ (marked up with <FAQPage> schema) increases the odds your site will be the source of the spoken response. Natural Language & Context: Write in a conversational tone where appropriate. Personal AIs use Natural Language Understanding to match user questions with answers. Content that reads like how a person would ask or answer a question performs better. For example, a blog title like “How can I improve my credit score quickly?” is more likely to align with a spoken query than a dry, keyword-stuffed title. In fact, content optimized for conversational context (long-tail, question-based queries) is now considered a best practice ( ^[17] ). Citations and Source Authority: AI agents will gravitate toward content that is factual, up-to-date, and authoritative . They may have internal confidence scores or cite sources for verification. Ensure your content is well-researched and cite reputable sources yourself. Being referenced by other trusted sites (digital PR for backlinks) also boosts the likelihood that AI will treat your content as part of its trusted corpus. By 2025, we’re seeing AI assistants referencing sources in responses more often for transparency. If your brand is a known authority in its domain (through consistent quality content and mentions across the web), personal AIs are more likely to include or recommend you when users ask for advice. Feed Personal Knowledge Graphs: Consider providing data to platforms that personal assistants use. For example, voice assistants often rely on knowledge partners – Alexa uses Yext/Wikipedia for general info, Siri uses sources like Yelp for local business info, etc. Ensuring your business listings, Wikipedia pages, or other knowledge entries are accurate can indirectly feed the answers these assistants give. In the future, as individuals use personal AI that learn from their data, even something like providing personalized content feeds or newsletter content that a user’s AI can ingest could make your brand part of that AI’s knowledge base. Forward-thinking companies are exploring offering plugins or integrations for AI platforms (similar to ChatGPT Plugins) so that personal AIs can fetch real-time data from their services when asked. In summary, personal AI assistants represent a shift from “pull” search (users querying search engines) to “push” or mediated search , where an AI intermediary pulls from various sources to deliver answers or actions. SEO in this realm is about being the source that these intermediaries trust and choose. As one AI SEO consultant noted, “By 2025, businesses that fail to prioritize AI SEO risk being left out of digital conversations and automated actions that matter most.” ( ^[18] ) Ensuring your content is AI-accessible, factually rich, and contextually relevant will position you to be included in the recommendations of tomorrow’s personal AI agents.

14.2 Multimodal Search and Answers: Beyond the Text Box

Traditional search has been dominated by text – users type words, and engines return text results. However, the next frontier is multimodal search , where images, voice, video, and even AR (augmented reality) content become part of both query and results. AI models like Google’s Gemini and OpenAI’s GPT-4 have been built from the ground up to be multimodal , meaning they can interpret and generate multiple forms of media. This opens up new possibilities: a user could snap a photo to search, or the search engine could respond with an image or interactive graphic rather than just text. Marketers will need to optimize content in ways that cater to all senses of search.

14.2.1 Gemini and the Era of Multimodal Models

– Google’s upcoming Gemini AI model exemplifies this trend. In 2025, Google began rolling out AI Mode in search (an experimental interface in Google Labs) that uses a custom Gemini model with advanced image understanding ( ^[19] ) ( ^[20] ). Users can now “search what you see” – for instance, by uploading a photo and asking questions about it ( ^[19] ). Google’s AI Mode can analyze the entire scene in an image: it recognizes objects, text, and context (materials, shapes, spatial relationships) at a sophisticated level ( ^[20] ). In one demo, a user snapped a photo of a bookshelf and asked, “Which of these books would be good for a 10-year-old who likes adventure?” The AI Mode identified each book spine, extracted their titles, queried reviews and age-appropriateness, and answered with a list of recommended books from that shelf, complete with purchase links – all in one step ( ^[21] ) ( ^[22] ). This is a dramatic expansion of search capability. Google’s Multimodal Search in action: On April 7, 2025, Google announced it had integrated Lens (its image recognition technology) into AI search results for a multimodal experience ( ^[23] ) ( ^[24] ). Early users found that AI Mode queries are on average twice as long as traditional searches ( ^[25] ) – suggesting people ask more detailed, conversational questions when interacting with a multimodal AI. They’re using it for tasks like comparing products, how-to guidance, and travel planning in a single query ( ^[26] ). This indicates that users treat the AI more like a knowledgeable assistant and less like a keyword-matching tool. The AI Mode doesn’t just show one image result: it performs a “query fan-out” , issuing multiple behind-the-scenes searches about different elements of the image to compile a comprehensive answer ( ^[27] ) ( ^[28] ). The result is an answer that may include text, images, and links blended together – e.g. a summary with illustrative images and clickable sources. Other players are also embracing multimodal answers. Bing Chat (with GPT-4) introduced image understanding in 2023 – users can upload a photo to Bing Chat and ask questions, and it will describe or analyze the image (for example, identifying a landmark from a photo) ( ^[29] ) ( ^[30] ). OpenAI’s ChatGPT Vision likewise can interpret images (e.g. you can show it a graph or a math problem picture and it will discuss it). On the output side, models can generate images as answers (OpenAI’s DALL·E integration, Bing’s Image Creator). We anticipate search results that include AI-generated visuals : imagine asking, “What did ancient Babylon look like?” and getting a AI-created reconstruction image alongside the explanation. Or a cooking search yielding an AI-generated image of the finished dish. Google’s Gemini is known to incorporate Google’s image generation model (Imagen), potentially to create images in responses where helpful ( ^[31] ).

14.2.2 Visual Search Optimization

– With images and multimedia becoming first-class citizens in search, SEO must expand to Visual SEO . This isn’t entirely new – Google Images and Pinterest have been popular for years, and Google Lens could already identify objects in photos – but the integration with conversational AI changes how visual results are used. Now, an AI might choose an image from your site to include or cite in an answer. Or it might prefer content that comes with a helpful diagram or infographic it can summarize. Key tactics for marketers include: High-Quality, Descriptive Images: Ensure you have clear, high-resolution images of your products or subject matter. AI vision models appreciate detail – a sharp image from multiple angles is more likely to be used or cited by an AI summarizer than a blurry one. According to SEO trend analyses for 2025, businesses using multiple high-quality images (with various angles/backgrounds) see better visibility in visual searches ( ^[32] ) ( ^[33] ). For example, an e-commerce furniture site should provide images of a chair from front, side, and in-context in a room; a travel blog should include photos of the destination landscape as well as maps. Alt Text and Metadata: Write descriptive alt text for every important image. Alt text serves two purposes: accessibility for users (screen readers) and a description that image-recognition AIs can latch onto. For instance, alt text like "alt='red vintage armchair with wooden legs in a modern living room setting'" gives the AI context it might not extract from the image alone (color, style, setting). Also use descriptive file names (e.g. red-vintage-armchair.jpg instead of IMG_1234.jpg ). Captions can be useful too; they are often given weight in understanding an image’s content. Structured Data for Images: Leverage schema markup such as ImageObject or product schemas that include image URLs. Google has been pushing for schema to enable features like product image search and 3D/AR results. If you have 3D models or AR content (e.g. a GLB/USDZ model of your product), use <script type="application/ld+json"> to mark it up (Google’s Search gallery provides guidelines for AR content schema). As multimodal search results start integrating AR, having a 3D model could mean your product can appear in a “View in AR” result in an AI answer. In fact, content with AR components may receive preferential treatment in certain search categories by 2025 ( ^[34] ) – for example, a retailer that offers an AR “try it in your room” feature might get highlighted in search over a competitor that doesn’t. Visual Content Uniqueness: Just as text content needs originality, so do visuals. If everyone is using the same stock photo of “call center customer service” on their webpage, an AI isn’t likely to choose that to display – it might pick none at all. Creating original graphics, charts, and photos not only engages users but also makes your content stand out to AI. For instance, if you publish a study , include a custom chart or infographic of the key findings. An AI summary might actually embed or generate a similar chart when answering a query about that topic, potentially citing your site as the source ( ^[35] ). Multimodal Content Strategy: Think beyond text when creating content. Could a tutorial be better as a short video or an interactive diagram? Google’s generative search can summarize videos too (e.g. Google has features that summarize YouTube videos, and future AI Overviews might integrate that). By providing content in multiple formats (text article, video, image gallery), you cover all bases for how an answer might be constructed. For example, a “how to fix a leaky faucet” guide is more GEO-optimized if it has a step-by-step text (for LLMs to parse), clear photos of each step (for visual search and possibly to be compiled into an AI-created montage), and maybe a short video (for users who ask their assistant to play a how-to).

14.2.3 Beyond Text: Video, AR, and Interactive Answers

– The multimodal future isn’t limited to static images. Video answers and augmented reality overlays are on the horizon: Video Summaries: You may soon ask an AI, “Show me how to tie a bowline knot,” and instead of linking to a YouTube video, the AI could generate a short video clip demonstrating the knot (using generative AI) or stitch together frames from existing videos. Already, Google’s SGE sometimes provides motion graphics or GIFs in answers for “how to” queries (e.g., showing a few frames of a process). Marketers should thus optimize video content as well, providing concise, well-structured videos with chapters. Ensuring your video content (on YouTube or your site) has transcripts and chapter markers helps AI identify relevant snippets. Also, be mindful that AI-driven video summaries might reduce the need for users to watch the full video, similar to how featured snippets reduced clicks – so consider watermarking your videos or mentioning your brand in the audio, so if the AI shows a portion, your brand is still evident. Augmented Reality Search: AR is set to make search more immersive. Google has already enabled AR search results for certain queries (e.g. viewing a 3D model of an animal or a retail product in your space via your phone). The expectation is that with devices like Apple’s Vision Pro and advanced smartphone AR, users might perform searches like “What is this building I’m looking at?” by just gazing at it through AR glasses, or “Are there any promotions for this product?” by scanning a product box in-store. Search will extend into physical space. For marketers, this means local and physical SEO converge with digital: ensure your physical locations are well-represented in digital maps and AR directories. Techniques like adding AR cues (e.g. scannable QR codes or AR targets on packaging that trigger content) can be part of marketing. Already, simple AR implementations can improve engagement and even rankings , as they tend to keep users on your site longer and interacting (a positive user signal) ( ^[36] ) ( ^[37] ). If your e-commerce site allows users to virtually “try on” a product via AR, not only can it boost sales, it may also be favored in search showcases for shopping searches. Interactive Result Formats: As AI synthesizes answers, we might see hybrid results – e.g. an answer with an embedded map for a local query, or a mini-app. Bing’s chat, for instance, can output interactive elements like clickable cards. Google’s AI Overviews already include interactive widgets for some shopping queries (like a carousel of products). Marketers should be aware of new structured data types and search features. For example, Google introduced things like “Circle to Search” (a Labs experiment allowing users to draw a circle around part of an image to refine a search) – your content’s images should be detailed enough to answer such drilled-down queries. Google is also working on multimodal search in Maps (point your camera at a street and search for restaurants that are highly rated on that street). Ensuring consistency between your website content and other representations (like Google Business Profiles, Google Maps info, etc.) is critical; contradictions (like different hours or pricing) could cause an AI to exclude or downrank you due to uncertainty. In essence, GEO in the multimodal age means treating images and visual experience as integral to SEO, not an afterthought. A 2025 SEO survey put it succinctly: “Visual search has come into its own… changing how people discover products and services” ( ^[38] ). Marketers should invest in visual content creation and optimization with the same rigor once reserved for text. By doing so, you increase your chances of being the source or inspiration for that rich, blended AI answer – whether it’s your image that’s displayed or your data powering an AR visualization. An example of voice-enabled AI search: Amazon’s Alexa, now enhanced with generative AI, can answer complex user queries conversationally. Optimizing content for such AI assistants (through structured data and authoritative information) will be crucial as users shift to multimodal and voice interactions. ( ^[10] ) ( ^[12] )

14.3 Voice Search 2.0: The Comeback of Conversation

Not long ago, voice search was heralded as “the next big thing,” and while adoption grew (Siri, Alexa, Google Assistant in millions of phones and homes), it never fully replaced typing for most users. One reason: voice assistants often gave underwhelming answers or just read snippets. However, with the advent of LLM-driven conversational AI, voice-based search is poised for a renaissance . We call this Voice Search 2.0 – more conversational, context-aware, and useful than the voice queries of the past.

14.3.1 From Queries to Conversations

– The biggest change is that voice interfaces can now engage in multi-turn conversations instead of one-shot Q&A. This dramatically improves their utility. For example, in the past you might ask Alexa a question and get a single answer or web result readout. Now, with generative AI, you can have a back-and-forth dialogue: “Alexa, plan a weekend trip to wine country.” It might respond with a summary of options, then you follow up, “I prefer something family-friendly and under $1,000 budget.” The assistant can dynamically refine its suggestions. This conversational capability means users are more likely to use voice for complex searches they avoided before. Indeed, data shows voice assistant users are becoming power users: globally, 32% of consumers used a voice assistant in the past week , and importantly 21% used it to find information (not just simple tasks like setting timers) ( ^[39] ). Usage of major voice tools has surged – Google Assistant usage jumped 46% in the U.S. between 2020 and 2024, and Siri by 40% ( ^[40] ). This growth correlates with the improvement in AI understanding and responding. A telling statistic: Nearly 1 in 3 voice assistant users have used ChatGPT in the past month ( ^[41] ). This indicates that people comfortable with voice are eagerly adopting advanced AI chat. Many have effectively turned ChatGPT into a voice assistant by using the mobile app’s voice input/output features. OpenAI’s release of whisper-level speech recognition and natural voice synthesis in ChatGPT (voice mode) in late 2023 made talking to ChatGPT feel like a human-like conversation ( ^[42] ). In other words, the line between a “search query” and a “conversation” is blurring . Users might start asking one question, then organically ask follow-ups as they would to a human expert, rather than doing separate keyword searches.

14.3.2 Voice Assistants Get Smarter (and More Popular)

– The big tech companies are racing to upgrade their voice assistants with LLM brains: Amazon Alexa’s AI Overhaul: As noted, Amazon is integrating Anthropic’s Claude into Alexa ( ^[10] ). The new Alexa (in preview as of 2024) can handle open-ended questions and multi-step requests. Amazon found the old Alexa’s limitations (needing exact phrases, inability to hold context) frustrated users, so this upgrade is aimed at making Alexa truly conversational. They even plan to monetize it (a premium subscription), which suggests they see enough value-add. Alexa’s active user base, once stagnating, is expected to grow again if the AI upgrade delivers. (eMarketer projects US voice assistant users will grow from 145 million in 2023 to 170 million by 2028 ( ^[43] ), modest but steady increases possibly boosted by improved AI functionality.) Google Assistant with Bard/Gemini: Google announced Assistant with Bard , essentially merging its voice assistant with generative AI ( ^[44] ). By early 2024, they began allowing users to opt Google Assistant on mobile to use Gemini (or Bard) as the backend ( ^[5] ) ( ^[45] ). This has unlocked new features: the Assistant can accept images and voice in the same query, and provide rich answers. A user might verbally ask, “Hey Google, what’s the name of this painting?” while pointing their phone camera at it – the Assistant (via the multimodal Gemini) can analyze the image and respond via voice: “This is ‘Starry Night’ by Vincent van Gogh.” It can also read long texts or web pages and summarize them aloud on command ( ^[6] ). Essentially, Google Assistant is becoming an accessible voice front-end for Google’s entire generative search power. Apple Siri and Others: Apple has been more secretive, but reports suggest they are working on an “LLM Siri” by 2026 ( ^[46] ). In the meantime, independent apps have brought AI to Apple devices (e.g. the “Perplexity AI” app on iPhone provides a voice Q&A interface with cited answers ( ^[47] ), effectively a third-party Siri alternative). Given Apple’s large installed base, once they integrate a powerful LLM, a huge number of users could suddenly have a vastly improved Siri. There are also regional players: in China, Baidu’s ERNIE Bot (a ChatGPT-like model) has voice input/output and is integrated into Baidu’s search and Xiaodu smart speakers; in India, where dozens of languages are spoken, local startups are making AI voice assistants tailored to vernacular languages – an important consideration for global SEO (content in local languages, and optimizing for voice transliteration and pronunciation). All these developments mean voice is becoming a preferred interface for many scenarios. It’s hands-free, convenient, and now – with better AI – actually helpful. A significant accessibility benefit drives this too: voice tech empowers those with visual or motor impairments to use the web. About 1 in 3 consumers with a visual impairment use voice assistants weekly , and many cite it as a crucial tool for independence ( ^[48] ). This underscores that optimizing for voice isn’t just about tech trends, but also about reaching users in the mode they find easiest.

14.3.3 Optimizing for Spoken Queries and Answers

– How does GEO adapt to voice 2.0? We build on the principles of Answer Engine Optimization (AEO) – delivering concise, direct answers – and tailor them to conversational contexts: Conversational Keyword Research: Voice queries tend to be longer and more natural language than typed queries ( ^[49] ). Instead of “weather Athens tomorrow”, a user will ask, “What will the weather be like in Athens tomorrow afternoon?” Capture these long-tail, question-format queries. Use tools or search query data to find common questions (especially beginning with who/what/how/where). Incorporate those questions verbatim in your content, and answer immediately after. For example, an FAQ entry: Q: “How long does shipping take to Europe?” A: “Shipping to Europe typically takes 5-7 business days.” This positioning makes it easy for an AI to grab the exact answer when someone asks their Echo device the same question. Featured Snippet Style Answers: Aim to get the featured snippet for key questions, as that often is what voice assistants read aloud. This means providing a concise answer in the first 1-2 sentences , then additional detail. Google’s voice answers usually truncate after ~30 seconds of speech (around 2-3 sentences). Structure your content accordingly. Also, use lists or steps for how-tos – voice AIs often enumerate steps one by one if they detect a list. For instance, a recipe site could include: “ Steps to make pancakes: (1) Mix flour and sugar... (2) Add milk... etc.” A voice assistant can then read the steps sequentially if asked. Optimize for “Speakability”: Writing for voice requires a slightly different tone and clarity. Use simple, clear language that’s easy to pronounce. Avoid jargon or complex sentence structures in the portions of text likely to be read aloud. Tools like Google’s Speakable schema (beta) allow you to explicitly markup parts of your content as optimized for TTS (text-to-speech). For news publishers, this schema can signal to Assistant which snippet to read. Even if you don’t use the markup, think in terms of how your content sounds . Read it out loud: does it sound like natural speech? Contractions (don’t vs do not), for example, can make content sound more conversational. Numeric information might need clarifying words (e.g. write “100 billion (100,000,000,000)” if you expect it to be spoken, so the user isn’t confused hearing “one hundred thousand million”). Local and Contextual Triggers: A huge portion of voice searches are local (“near me” queries, or location-based) and transactional (calling a business, ordering something). Ensure your Google Business Profile is up to date, as Google Assistant heavily relies on that for local questions ( “Find me a nearby coffee shop open now” will check Google Maps data). Also consider that voice often comes from mobile on-the-go – optimize for mobile and fast load, since sometimes the assistant might hand off to your site on the phone (if it can’t answer fully). Use schema for local business , including hours, addresses – this helps voice assistants provide direct answers (“The store is open until 9 PM”). Integration with Voice Platforms: Think about creating content specifically for voice platforms. For example, Alexa skills or Google Assistant actions – a few years ago these were niche, but with new AI capabilities they could make a comeback with less development effort (because you can rely on the AI for natural language handling). An example might be a Flash Briefing if you’re a publisher (concise news summary that Alexa can read to users). Or a Q&A database integration: e.g. a healthcare company might publish a set of health Q&As that voice assistants can draw from when users ask health questions (some organizations partnered with Alexa to provide trusted answers). While not every brand needs a custom voice app, exposing your content via APIs or feeds that assistants use (such as RSS for news, or being part of voice commerce programs) can give an edge. Measure and Adapt: Analytics for voice are notoriously sparse (you often don’t know what was asked or if your answer was used). However, you can gather some signals: monitor the questions that lead users to your site (in Google Search Console, filter queries that are phrased as questions – these often originate from voice). Also watch metrics like zero-click searches and branded voice mentions. Some third-party tools and new features are emerging to track AI and voice visibility (see Chapter 13 on measuring AI references). User surveys or feedback can also help – ask customers if they found you via a voice assistant. As voice search grows, more formal metrics may appear (Google might eventually show voice impressions in Search Console, etc., especially if regulatory transparency increases). In summary, Voice 2.0 means revival with a twist : it’s not the simplistic voice search of the late 2010s; it’s smarter and integrated with AI. Marketers who adapt content to be the answer spoken by these voice AIs will tap into a user base that is highly engaged. In fact, voice assistant users are a valuable segment – they tend to be younger, affluent, and spend more online (voice users are 33% more likely to have made an online purchase in the past week, as mentioned) ( ^[50] ). By making sure your content “speaks” well, you cater to this segment. As the GWI report on voice trends 2025 concludes: brands that act early on voice get “stronger discoverability, deeper engagement, and loyalty that lasts.” ( ^[51] ) Voice is becoming an instinctive way people interact with tech – optimizing for it ensures you remain part of those interactions as they become more frequent and natural.

14.4 Evolving AI Algorithms and Transparency: Navigating New Rules

As generative AI becomes deeply embedded in search, the algorithms that power search results – and the rules governing them – are rapidly evolving. There is a dual force at play: technological updates to AI models (improving quality, reducing hallucinations) and external pressures (regulators and publishers pushing for transparency, credit, and control). Marketers will need to stay on top of these changes to ensure their GEO strategies remain effective and compliant. In this section, we explore how AI search algorithms might change and the moves toward greater transparency and attribution in AI-driven results.

14.4.1 Continuous AI Model Upgrades

– Unlike the relatively gradual updates of classic search algorithms (e.g. Google’s core updates a few times a year), LLM-based systems can have major leaps with new model versions. For instance, OpenAI’s jump from GPT-3.5 to GPT-4 dramatically improved factual accuracy and reasoning. Google’s progression from Bard (based on LaMDA) to Gemini 2.5 in Search Labs brought multimodal capabilities and better understanding of complex queries ( ^[52] ) ( ^[53] ). We should expect frequent upgrades : Google has hinted at Gemini iterations, Microsoft/Bing will integrate OpenAI’s latest models (GPT-5 when available), and new entrants like xAI’s Grok are on version 4 already by mid-2025 ( ^[54] ) ( ^[55] ). What this means for GEO is that ranking factors for AI answers may shift . In traditional SEO, we obsess over hundreds of signals (links, keywords, Core Web Vitals, etc.), but for AI-generated answers, the signals might include: Content quality and accuracy (the AI has internal “scoring” for how reliable an extracted passage is, possibly based on source reputation or embedding similarity to known facts). Schema/structured data usage (to directly answer certain queries, the AI might favor sites with structured info it can easily parse). Popularity and reinforcement : If an AI model was trained on a snapshot of the web, content that was widely cited or linked as of that training cut-off could have a higher prior probability of being used in answers. However, with retrieval-augmented generation (RAG), current popularity (like top search results in real-time) matters too ( ^[56] ) ( ^[57] ). Terakeet’s research noted a strong correlation between a domain’s Google organic ranking and its likelihood of being included in Google’s AI Overviews ( ^[57] ). So traditional SEO success still boosts GEO success: if you rank well for a topic, Google’s AI is more likely to pull you in. Furthermore, AI answers can incorporate user behavior signals in new ways. For example, if users frequently click a particular citation in an AI answer and seem satisfied (no further follow-ups), the system could learn to prioritize that source for future similar queries. Google has alluded to using user feedback loops to refine SGE results – users can give a thumbs-up/down on AI answers, which could indirectly affect which sources are chosen. Marketers should be prepared that AI algorithms will be somewhat of a black box , but likely reward the same core principle as Google always has: content that best satisfies the user’s query . Now, however, satisfaction is measured by the AI’s ability to address the query, possibly synthesizing multiple sources.

14.4.2 Transparency, Citations, and Content Attribution

– A critical and contentious area is how AI search experiences credit content creators. Early on, systems like ChatGPT would generate answers with no citations, effectively absorbing content without attribution. This raised alarm among publishers. In response, some AI search engines differentiated by citing sources – for instance, Perplexity AI built its brand on providing footnoted answers with references ( ^[47] ). Microsoft’s Bing Chat also from launch would include citations (though not always comprehensively). Google’s SGE initially showed sources in small card links, but not clearly mapping each sentence to a source, which drew criticism ( ^[58] ). By 2024-2025, regulatory and public pressure have grown for more transparency. The EU’s pending AI Act (expected to be in force by 2025) includes provisions about disclosure of AI-generated content and respecting copyrights ( ^[59] ) ( ^[60] ). There’s talk that generative search outputs, especially for news, may need to clearly cite sources to avoid misinformation and support the journalism industry. In fact, Google decided against requiring publisher opt-in for AI snippets , drawing industry ire ( ^[61] ). Major news organizations (through bodies like the News/Media Alliance) have petitioned regulators, arguing that AI search “misappropriates” their investment in content ( ^[62] ). All this indicates the trend is toward more visible citations and possibly opt-in systems . Google has signaled that it wants to “prioritize traffic to websites and merchants” even as it rolls out AI overviews ( ^[63] ). It shared data that in early testing, people clicked links in AI Overviews and those clicks were actually higher quality (users spent more time on site) ( ^[64] ) ( ^[65] ). This positive spin suggests Google is trying to reassure publishers that AI results won’t destroy their traffic. Indeed, Google claimed that when people click AI results links, they are more likely to be satisfied and engage longer ( ^[64] ). As GEO strategists, we should track such metrics (e.g., if you do get traffic from an AI answer, are those users more engaged? Possibly because the AI primed them well). Practically, we expect search UIs to experiment with showing sources more clearly. Bing, for example, might move to an “inline citation” model (like academic papers, little reference numbers next to facts). Google’s SGE could start highlighting which part of the answer came from which URL on hover or click. For marketers, this means earning that citation is the new click-through . Your content might be seen (or heard) by millions via AI answers, but you may get fewer clicks – so the branding and messaging within the answer itself becomes crucial. Ensure your brand name or a relevant mention is in the snippet that AI often uses. For instance, an AI might quote: “According to Acme Insurance’s site , the average cost of car insurance in 2025 is $X.” If your site explicitly says “According to Acme’s 2025 Cost Index, ...” you have a better chance that the AI includes those words and gives you a verbal mention, not just a footnote. This is a subtle but important point: structure your sentences in a way that if excerpted, they carry attribution. (Many publishers now insert phrases like “Our research at [Company] shows…” to prompt citation of their name.) We may also see the rise of content agreements : already, some publishers have licensed content to AI companies (OpenAI licensed a corpus from the AP, for example). If such deals proliferate, content behind paywalls or from premium providers might start appearing in AI answers (with permission). Marketers might consider syndication deals or partnerships so that their content is part of premium data sets feeding personal assistants or vertical AI (imagine a travel AI that has licensed Lonely Planet content – a hotel that wants to be mentioned might need to ensure it’s listed or advertising in those sources).

14.4.3 Content Control: Opt-Out and New Meta Tags

– With AI scraping the web, site owners sought ways to opt out if desired. In mid-2023, OpenAI introduced a GPTBot crawler and allowed sites to block it via robots.txt directives, and Google’s Bard did similar (using User-agent: Google-Extended ) ( ^[66] ) ( ^[67] ). By adding a couple lines to robots.txt , sites can request not to be used for AI training ( ^[68] ). However, these measures apply only to training data, not necessarily to on-the-fly retrieval. Still, they were an important first step in giving publishers a voice. Many major sites (CNN, Reuters, Wikipedia, etc.) put up such blocks until they get compensation. Moving forward, we might get more granular controls. For example, a meta tag like <meta name="ai:allow" content="no"> could emerge to signal “do not include my content in AI answers.” If regulators push for it, search engines might honor such tags in generative experiences. Alternatively, tags to encourage citation: e.g. <meta name="ai:citation_title" content="Acme Research"/> to tell the AI how you want your source to be named if cited (this is speculative, but in line with how meta descriptions guide snippet text in regular search). Publishers need to weigh opt-out carefully. Opting out entirely (as some did from training) could mean losing visibility if AI search overtakes traditional. It’s a double-edged sword: opt-in might yield credit but fewer clicks; opt-out might preserve clicks for now but risk invisibility in new interfaces. The consensus in the SEO community is to stay visible but adapt measurement . Instead of chasing only clicks, monitor mentions in AI answers , brand impressions, and downstream actions (e.g. someone heard your brand via voice AI and later Googled it or navigated directly – not easy to track, but brand lift surveys could help).

14.4.4 Regulatory Environment and Standards

– Globally, regulations will influence GEO. The EU AI Act (likely effective 2025) will require transparency for AI systems, especially in high-risk areas (like medical or financial advice) ( ^[59] ) ( ^[69] ). Search might not be classified as high-risk, but any AI-generated content might need labeling. We could see disclaimers like “(AI-generated)” on search answers. The act also has a provision that AI models must document their training data sources to some extent. This could eventually pressure companies to offer an API or tool for webmasters: e.g. “Was my content used in your training data?” (OpenAI actually launched a tool for journalists to check if their articles were in GPT-4’s training set). Marketers might gain insights into how their content is being used by AI. In the US, discussions at the FTC have hinted at scrutinizing if using content without permission in AI outputs is an antitrust or unfair practice – especially since Google both controls the search traffic and now uses content in AI form (publishers argue this is a conflict). In Asia, China’s regulations already require AI chatbots to cite sources for news-related queries to curb misinformation. So a likely outcome is an industry-wide standard for AI citation . There might be guidelines or best practices that major AI providers agree to, in order to appease publishers and regulators, such as: always provide at least three source links for factual answers , or if more than 50% of an answer’s wording comes from one source, cite it explicitly , etc. For GEO, this means you should welcome citation opportunities . For instance, if you run a Q&A forum or UGC site, you might structure your licensing to allow AI usage with attribution, thereby becoming a source of answers (StackExchange did something like this, allowing training use as long as there’s a hyperlink back). This could drive indirect traffic or at least brand awareness.

14.4.5 Preparing for Algorithmic Shifts

– Finally, marketers must remain agile. As AI algorithms update, watch for changes in how your content surfaces. Some practical tips: Monitor AI answer presence: If possible, use tools that simulate queries in SGE, Bing Chat, etc., and see if/where you appear. There are startups now offering “AI result monitoring” that track if a brand is mentioned by AI for certain queries. Quality & E-E-A-T: Every trend points to quality content being even more paramount. Google integrated its Helpful Content system into core ranking in 2024 and explicitly trained it to penalize AI-generated, unoriginal content ( ^[70] ) ( ^[71] ). Its Search Quality Rater Guidelines as of Jan 2025 say to flag AI-mass-produced, unoriginal content as lowest quality ( ^[72] ). This is no coincidence – it’s a direct response to the AI content flood. So, avoid any temptation to churn out lots of mediocre AI-written pages; focus on unique insights and depth (we’ll expand on this in 14.5). Meta signals for AI: Keep an eye on any new schema or meta tags related to AI. For example, Google recently encouraged use of indexifembedded (for controlling if content can be used in embeddings like Bard). Use these as they fit your strategy. Stay informed: Make it a habit to follow updates from search engines about their AI experiments. Google, for instance, posts on its blog about SGE changes (like adding videos in answers, or expanding to new countries). These updates can hint at what they prioritize. If Google says “we’re experimenting with citing every sentence’s source ,” you know to ensure your sentences stand strongly on their own with context. In summary, the rules of engagement in AI search are being rewritten in real-time. Transparency and fairness are the key themes: Which content is used, how it’s credited, and whether creators have a say. For marketers, adaptation is two-fold: adjust your tactics to align with algorithmic changes (e.g. double down on structured data, quality, and making your content easy for AI to use accurately), and adjust your expectations and KPIs (views and mentions may supplement clicks as measures of success). By navigating these evolving rules proactively, you can maintain and even grow your visibility as search transitions into this new AI-driven paradigm.

14.5 The Enduring Role of Human Creativity: A Competitive Edge

In a world where AI can generate endless content, one might fear that human-produced material will be drowned out. Indeed, by some estimates, 90% of online content could be synthetically generated by 2026 ( ^[73] ). This tsunami of AI-generated text, images, and videos raises the question: how will search engines and users separate the wheat from the chaff? The answer increasingly comes down to human creativity, experience, and originality . Paradoxically, the more AI content proliferates, the more valuable truly human content becomes – both to search algorithms seeking quality and to users seeking authenticity.

14.5.1 Content Deluge and the Need for Originality

– We are already seeing the effects of mass AI content. Marketers armed with GPT-4 have pumped out blog posts at an unprecedented scale (content generation up 8,000% since 2022 by one account ( ^[74] )). The result is an “AI echo chamber,” where many articles on the web repeat the same mediocre insights, because they were all spun from the same training data. This poses a big problem: if everyone’s using similar AI models, a lot of content starts sounding the same – “beige,” as one commentator put it ( ^[75] ). Users notice this. Search engines definitely notice this. Google’s 2024 core update was explicitly aimed at demoting unoriginal content; originality, depth, and expertise became make-or-break ranking signals ( ^[76] ). This trend will only intensify. Google and other search engines are deploying ever more sophisticated quality filters. As of early 2025, Google’s quality raters (who help train the algorithms) were instructed to give the lowest quality rating to “content that appears to be largely AI-generated with no original contribution” ( ^[72] ) ( ^[71] ). In other words, thin, regurgitated content is being aggressively weeded out . The Helpful Content system, now part of the core algorithm, uses machine learning to identify patterns of mass-produced content. Sites that, for example, publish dozens of similar articles with AI and no unique point of view have seen drops in visibility. Some affiliate sites experienced this in late 2023 – they let AI write dozens of product “reviews” that were generic, and Google’s updates crushed their rankings.

14.5.2 E-E-A-T and Trust as Differentiators

– Google’s concept of E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) has become the guiding light for content quality. AI can mimic expertise to a degree (scraping facts from the web), but it lacks real-world experience and insight. It cannot testify to using a product, visiting a place, or having a professional credential (unless provided by a human in training data). That’s why Google added the extra “E” (Experience) to E-A-T in late 2022 – to value content produced from first-hand experience. For marketers, this means highlighting human perspective. Use authors with credentials and bios. Share personal anecdotes or case studies that an AI wouldn’t have. For example, a travel blog that shares “I hiked the Inca Trail in March 2024 – here’s what I learned” has inherent experience that an AI writing “Top 10 hikes in Peru” lacks. We see evidence that content with author authority is being favored. LinkedIn, for instance, did an analysis of search results in the finance niche after Google’s 2024 updates: sites that showcased author profiles with financial certifications and personal insights saw improved ranking, whereas faceless “SEO content” sites fell. On Google’s part, they even experimented with labeling some search results with “Perspectives” (to highlight forum posts or firsthand accounts) in contrast to generic AI-scraped info. And remember, AI itself, when uncertain, might choose to quote a human expert. For instance, an AI answer about medical advice might cite a line like “ Dr. Jane Smith, a board-certified dermatologist, explains that… ” if such a quote is available. Thus, if you have expert staff, incorporate their quotes and names into your content – it can both improve your credibility and create hooks that AI might use. Trustworthiness also involves accuracy and honesty . AI content can sometimes be confidently wrong (so-called hallucinations). A human touch – citing sources, providing transparency (“ Data as of Q1 2025… ”), acknowledging uncertainties – can make content more trustworthy. Brands that double-down on factual accuracy (even hiring fact-checkers for content) will stand out as reliable in an AI-saturated web.

14.5.3 Creativity and Brand Voice

– When everyone is using AI to write similarly, injecting creativity and a strong brand voice sets you apart. Consider Wendy’s fast-food chain on social media – their snarky, human tone made them famous; an AI could copy their style only after Wendy’s established it. Original research, unique opinions, compelling storytelling – these human elements create memorable content . AI by design tends to average out to the mean of what’s in its training. It’s not innovative ; it’s derivative. So, use that to your advantage: dare to present insights or angles that haven’t been said a thousand times. Conduct small experiments or surveys and publish results (now your content is one-of-a-kind data). Tell stories or incorporate user-generated content – these add unpredictability that AI can’t fabricate easily. Not only will this help in rankings, it will make your content more likely to be referenced by others (human or AI). An AI might skip over a generic paragraph, but a unique statistic or a compelling anecdote might survive in the synthesized answer because it adds value. For example, if your fitness blog includes a personal story of overcoming injury that resonates, an AI might include a snippet (“ One marathon runner on FitLife blog mentioned she recovered in 8 weeks by… ”) as a way to humanize an answer about injury recovery.

14.5.4 Human-AI Synergy

– Embracing human creativity doesn’t mean rejecting AI tools. The future competitive edge is combining the two wisely. Use AI for what it excels at: data crunching, drafting routine sections, translating, etc., but always add a human layer on top. Think of AI as the junior copywriter and your human team as editors and strategists. This can even be communicated transparently, which can build trust (“ This article was fact-checked by our editorial team and informed by AI analysis of 1,000 reviews ”). Users will increasingly appreciate knowing content’s provenance – being clear that a human curated or wrote it (or at least verified it) can become a selling point. We should note that search algorithms are likely to develop AI-content detection signals – not outright “AI detectors” (which are unreliable), but proxies: e.g. if a site’s content all reads at a formulaic 8th-grade level with no author bios and no unique info, it’s a sign of mass production. On the flip side, if a site consistently publishes original research, gets mentioned by others, and has engaged readership, it signals quality. The enduring SEO playbook, thus, circles back to quality over quantity . It’s better to have one outstanding, link-worthy, case-study-filled article than 50 generic listicles generated by an AI. This was always somewhat true, but now it’s absolutely crucial.

14.5.5 Examples of Human-Centric Content Wins

– To illustrate, consider Red Bull’s content strategy . Red Bull focuses on extreme sports and adventure storytelling aligned with their brand. They produce high-quality documentaries, athlete interviews, and event coverage that no AI could replicate without being given the footage. The result: their content is highly shareable and often surfaces in search for topics like “freestyle mountain biking highlights” – because it’s authentic and unique. In the B2B space, a company like HubSpot has thrived by publishing original research (marketing data reports) and thought leadership by named experts. Even though countless AI-written marketing tips articles exist, HubSpot’s pieces continue to rank well and be cited, in part due to that authority and originality. On the other hand, sites that tried to game the system with AI content have seen backlash. In mid-2023, a travel site that auto-generated city guides with AI had to pull them down after readers found they contained factual errors and banal info. Google’s Helpful Content Update hit them, and their traffic dropped. The lesson: unoriginal AI content can actively hurt your SEO – Google will not only ignore it, but penalize the domain if it becomes prevalent ( ^[70] ) ( ^[71] ). Recovery requires pruning low-value pages and improving what remains with real human input. To sum up, human creativity, experience, and insight are your competitive moat in the era of AI. As generative models make basic content a commodity, what stands out is the genuine and the novel . In practical terms for GEO: Invest in content that showcases real expertise or experience (authoritative whitepapers, personal stories, expert interviews). Foster a distinct brand voice and community – things that create an emotional connection or loyalty, which no generic content can steal. (This can indirectly boost SEO through branded searches and mentions.) Continuously update and refine content to add new human insights. An AI might not update its training data often; if you’re the first with a new insight in 2025, you become the source that all the AI-driven sites cite. Finally, recall that search engines themselves ultimately want to satisfy users. If AI content flood leads to lower user satisfaction, the algorithms will adjust to favor content that does satisfy – which will be the more creative, accurate, and trustworthy human content. In an ironic twist, the age of AI could reorient digital marketing away from volume and back toward fundamentals of creativity, quality, and authenticity . For those willing to uphold those values, the future is bright: you’ll rise above the noise (human and machine-made) as a beacon of reliable, interesting information. As one digital strategist quipped, “In the era of infinite content, being real is the ultimate differentiation.”

References

^[1] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[2] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[3] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[4] Blogs.Microsoft.Com Article - Blogs.Microsoft.Com URL: https://blogs.microsoft.com/blog/2024/11/19/ignite-2024-why-nearly-70-of-the-fortune-500-now-use-microsoft-365-copilot

^[5] TechCrunch Article - TechCrunch URL: https://techcrunch.com/2024/02/08/google-assistant-is-now-powered-by-gemini-sort-of

^[6] TechCrunch Article - TechCrunch URL: https://techcrunch.com/2024/02/08/google-assistant-is-now-powered-by-gemini-sort-of

^[7] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[8] Blogs.Microsoft.Com Article - Blogs.Microsoft.Com URL: https://blogs.microsoft.com/blog/2024/11/19/ignite-2024-why-nearly-70-of-the-fortune-500-now-use-microsoft-365-copilot

^[9] Blogs.Microsoft.Com Article - Blogs.Microsoft.Com URL: https://blogs.microsoft.com/blog/2024/11/19/ignite-2024-why-nearly-70-of-the-fortune-500-now-use-microsoft-365-copilot

^[10] www.reuters.com - Reuters URL: https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30

^[11] www.reuters.com - Reuters URL: https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30

^[12] www.reuters.com - Reuters URL: https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30

^[13] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[14] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[15] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[16] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[17] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[18] Xponent21.Com Article - Xponent21.Com URL: https://xponent21.com/insights/ai-seo-strategies-for-2025

^[19] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[20] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[21] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[22] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[23] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[24] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[25] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[26] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[27] Google Article - Google URL: https://blog.google/products/search/ai-mode-multimodal-search

^[28] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/644363/google-search-ai-mode-multimodal-lens-image-recognition

^[29] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/644363/google-search-ai-mode-multimodal-lens-image-recognition

^[30] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/644363/google-search-ai-mode-multimodal-lens-image-recognition

^[31] TechCrunch Article - TechCrunch URL: https://techcrunch.com/2024/02/08/google-assistant-is-now-powered-by-gemini-sort-of

^[32] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[33] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[34] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[35] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/644363/google-search-ai-mode-multimodal-lens-image-recognition

^[36] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[37] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[38] www.theedigital.com - Theedigital.Com URL: https://www.theedigital.com/blog/seo-trends-2025

^[39] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[40] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[41] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[42] OpenAI Article - OpenAI URL: https://openai.com/index/chatgpt-can-now-see-hear-and-speak

^[43] www.emarketer.com - Emarketer.Com URL: https://www.emarketer.com/content/voice-assistant-user-forecast-2024

^[44] 9To5Google.Com Article - 9To5Google.Com URL: https://9to5google.com/2024/01/01/google-bard-2024-features

^[45] TechCrunch Article - TechCrunch URL: https://techcrunch.com/2024/02/08/google-assistant-is-now-powered-by-gemini-sort-of

^[46] Em360Tech.Com Article - Em360Tech.Com URL: https://em360tech.com/tech-articles/apple-ai-upgrade-launch-llm-siri

^[47] www.theverge.com - Theverge.Com URL: https://www.theverge.com/news/644363/google-search-ai-mode-multimodal-lens-image-recognition

^[48] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[49] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[50] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[51] www.gwi.com - Gwi.Com URL: https://www.gwi.com/blog/voice-search-trends

^[52] Search.Google Article - Search.Google URL: https://search.google/ways-to-search/ai-mode

^[53] Search.Google Article - Search.Google URL: https://search.google/ways-to-search/ai-mode

^[54] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[55] www.businessinsider.com - Businessinsider.Com URL: https://www.businessinsider.com/grok-artificial-intelligence-chatbot-elon-musk-xai-explained-2025-7

^[56] Builtin.Com Article - Builtin.Com URL: https://builtin.com/articles/generative-engine-optimization-new-seo

^[57] Builtin.Com Article - Builtin.Com URL: https://builtin.com/articles/generative-engine-optimization-new-seo

^[58] www.washingtonpost.com - Washingtonpost.Com URL: https://www.washingtonpost.com/technology/2024/05/13/google-ai-search-io-sge

^[59] Digital-Strategy.Ec.Europa.Eu Article - Digital-Strategy.Ec.Europa.Eu URL: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai

^[60] www.deloitte.com - Deloitte.Com URL: https://www.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2024/tmt-predictions-eu-generative-ai-regulation.html

^[61] www.bloomberg.com - Bloomberg.Com URL: https://www.bloomberg.com/news/articles/2025-05-19/google-gave-sites-little-choice-in-using-data-for-ai-search

^[62] www.newsmediaalliance.org - Newsmediaalliance.Org URL: https://www.newsmediaalliance.org/release-news-media-alliance-calls-on-ftc-doj-to-investigate-googles-misappropriation-of-digital-news-publishing-stop-expansion-of-generative-ai-overviews-offering

^[63] www.washingtonpost.com - Washingtonpost.Com URL: https://www.washingtonpost.com/technology/2024/05/13/google-ai-search-io-sge

^[64] Google Article - Google URL: https://blog.google/products/ads-commerce/ai-creativity-google-marketing-live

^[65] Google Article - Google URL: https://blog.google/products/ads-commerce/ai-creativity-google-marketing-live

^[66] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[67] www.eff.org - Eff.Org URL: https://www.eff.org/deeplinks/2023/12/no-robotstxt-how-ask-chatgpt-and-google-bard-not-use-your-website-training

^[68] Raptive.Com Article - Raptive.Com URL: https://raptive.com/blog/how-to-prevent-gptbot-from-crawling-your-site

^[69] www.lowenstein.com - Lowenstein.Com URL: https://www.lowenstein.com/news-insights/publications/client-alerts/the-eu-artificial-intelligence-act-of-2024-what-you-need-to-know-privacy

^[70] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

^[71] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

^[72] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

^[73] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

^[74] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

^[75] www.zoomsphere.com - Zoomsphere.Com URL: https://www.zoomsphere.com/blog/lets-talk-ai-content-creation-and-why-its-giving-seo-a-headache

15. Conclusion and Next Steps

15.1 Embracing the Change: From SEO to GEO

Figure: A comparison of generative AI-driven search traffic growth vs. traditional organic search (2024–2025). Recent research shows generative AI discovery skyrocketing – growing 165× faster than organic search traffic ( ^[1] ). This explosive rise signals a fundamental shift in how users find information online ( ^[2] ). Search marketing as we know it is undergoing a seismic transformation – but it is not dying. In fact, it’s evolving into something broader and more nuanced. The rise of generative AI in search represents a new chapter in this evolution. Marketers are witnessing search queries transition from simple keywords to complex, conversational prompts answered by AI. Those who embrace the change and adapt their strategies to include Generative Engine Optimization (GEO) will sustain and even grow their online visibility, while those clinging to old playbooks risk being left behind ( ^[3] ). As former IBM CEO Ginni Rometty aptly put it, “AI will not replace humans, but those who use AI will replace those who don’t.” ( ^[3] ) – a pointed reminder that adopting AI-driven tactics is now a competitive necessity. The journey from SEO to GEO has been a recurring story of adaptation. Ever since search engines debuted, SEO pundits have periodically proclaimed “SEO is dead,” only to see it reborn in a new form each time. We saw this when mobile search rose, when social media emerged, and now we see it with AI. Indeed, organic search traffic from Google is under pressure – roughly 64% of Google searches result in no click to external sites today ( ^[4] ), as users either get their answer directly on the results page or are siphoned into Google’s own properties. Rather than heralding the end of search marketing, this trend signals that search behavior is shifting to a model where answers are often provided instantaneously by AI. Marketers must respond by ensuring their content is present in those instant answers. As one 2025 industry analysis bluntly stated, “Google organic search traffic is gone and isn’t coming back… So keep doing legacy SEO – but evolve .” ( ^[5] ) The message is clear: classic SEO tactics alone will no longer suffice in this new era. Embracing GEO means recognizing that search engines are morphing into answer engines and conversation partners . Google’s Search Generative Experience (SGE) is a prime example. Instead of the familiar list of blue links, SGE uses AI to generate a rich snapshot answer at the top of the page, synthesizing information from various sources ( ^[6] ). Users can get the gist of a topic at a glance, then dive deeper via follow-up questions or cited links. In practice, this transforms the user experience dramatically. Figure 15.1 below shows an example of Google’s SGE in action: an AI-generated overview appears in a shaded box at the top of the results, answering a complex question, with sources listed to “learn more.” The traditional search results are pushed further down, and users are invited to continue the dialogue with additional questions ( ^[7] ) ( ^[8] ). Figure 15.1: Example of Google’s Search Generative Experience (SGE) interface. An AI-powered “snapshot” answer (top) provides a conversational response with cited sources, fundamentally changing how users interact with search results ( ^[6] ) ( ^[9] ). This evolution from “10 blue links” to AI-driven answers means search marketers must broaden their perspective. It’s not that people have stopped searching for information – far from it. It’s that how and where they search is fragmenting across new platforms and AI assistants. ChatGPT’s mainstream breakthrough in late 2022 familiarized millions with conversational search. By 2024, over 13 million U.S. adults had made generative AI (like ChatGPT) their go-to tool for online discovery ( ^[10] ). Microsoft’s Bing integrated GPT-4 into its search in early 2023, offering chat-based results alongside web links. Google launched SGE and later rolled out “AI overviews” by default in its results ( ^[11] ). New AI-driven search engines like Perplexity AI emerged, delivering direct answers with up-to-date citations instead of just links. In China, Baidu introduced its Ernie Bot to deliver AI answers in search, while local players in other regions (like Naver’s HyperCLOVA in Korea) began experimenting with generative search features. In short, search is no longer a monolith – it’s an ecosystem of AI assistants, chatbots, and traditional engines all coexisting. For marketing professionals, this is a challenging but exciting shift. It requires a dual mindset: continue to value the fundamentals that made SEO successful (relevance, quality content, technical soundness), while also innovating new techniques to ensure your content is preferred by AI models. The good news is that many SEO best practices still apply – perhaps more than ever. The core mission remains: connecting your audience with the information and solutions they seek. What’s changing is the medium through which that connection happens. Instead of solely optimizing for search engine algorithms and browsers, we now must optimize for AI models and chat interfaces as well. It’s a layer of optimization on top of SEO – not a replacement of it. Indeed, forward-thinking experts have started referring to this holistic approach as “search everywhere optimization,” meaning optimizing your presence wherever people (or their AI agents) might be seeking answers ( ^[12] ). In essence, GEO is about meeting your users wherever and however they search – be it a Google AI overview, a Bing chat answer, a voice query to Alexa’s new LLM-powered assistant, or a question posed to ChatGPT. Search marketing is alive and well; it’s just wearing a different outfit. Those who adapt will find that the new opportunities can augment their reach rather than diminish it. To sum up the journey from SEO to GEO: adaptation is survival . The past chapters have shown that search marketing isn’t ending – it’s expanding. By embracing generative AI and integrating it into your search strategy, you ensure your brand remains visible in the next generation of search. The marketers who view AI not as a threat but as an extension of their toolkit will thrive. Just as mobile-responsive websites, social media engagement, and voice search optimization became new arrows in the SEO quiver over the years, now GEO tactics join that arsenal. The companies that start blending SEO with GEO now are positioning themselves to capture both today’s search traffic and tomorrow’s AI-driven discoveries. In contrast, those who ignore this shift risk losing relevance as user behavior evolves. The choice is stark, but the path forward is inspiring: by staying true to the fundamental goal of helping users while leveraging cutting-edge AI technology, search marketers can navigate this evolution with confidence. As one industry veteran observed, “SEO isn’t going anywhere. It still matters and will continue to be a thing… But it’s impossible to ignore this new ‘thing’ [AI search]” ( ^[13] ). In other words, keep your SEO skills sharp – but be ready to reinvent and expand them for the AI era . Embracing the change is the first step toward sustained success in the era of Generative Engine Optimization.

15.2 Key Takeaways: SEO + GEO Synergy for Success

In this section, we distill the key lessons from our exploration of SEO and GEO. The takeaway is that success comes from blending the timeless principles of search optimization with new tactics for AI-driven platforms. By integrating the old and the new – classic SEO and cutting-edge GEO – marketers can maintain their visibility across all search channels. Below are the core principles to remember: Quality Content & Authoritativeness Are Paramount: Content is still king , perhaps even more so in the age of AI. Generative models like ChatGPT and Google’s SGE draw answers from content they deem authoritative and trustworthy . This means you must continue to create content that demonstrates expertise, experience, authority, and trustworthiness (E-E-A-T). Original research, in-depth analysis, and unique insights set your content apart from the generic noise that AI often generates. High-quality content not only ranks well in traditional search, but it also increases the likelihood of being referenced by AI answers. Remember, the **future is not about just ranking a page – it’s about having a brand so authoritative that it gets mentioned or cited by AI platforms ( ^[5] ). If an AI assistant trusts your site enough to quote it or recommend it, that’s the new gold standard. So double down on content that provides real value – answer the questions your audience is asking (especially the nuanced ones), and back up your answers with credible data and references. By doing so, you build the kind of digital reputation that both humans and algorithms recognize as authoritative. Technical Soundness & Structured Data: The technical foundations of your website remain critical in the GEO era. Search engines and AI crawlers need to efficiently access, understand, and extract information from your site. This means ensuring crawlability, indexability, and fast performance – all the traditional technical SEO must-do’s. Additionally, focus on structured data and semantic HTML to make your content AI-friendly. Schema markup (for FAQ, How-To, Product, etc.) provides context that can help generative AI understand and confidently use your content ( ^[14] ). For example, adding FAQ schema or clearly structured headings can increase the chances that an AI overview pulls in your text as a featured snippet or cited answer. Clean HTML structure (using proper heading tags, lists, tables) makes it easier for algorithms to parse your content. Also, site speed and mobile-friendliness continue to matter – not just for SEO rankings, but for user experience in scenarios where AI directs a user to your site. If someone clicks a link from an AI answer and your page loads slowly or is unreadable on mobile, you’ve lost the opportunity. In summary, an AI-ready site is one that follows best technical practices : fast, accessible, well-structured, and marked up with relevant metadata. Multi-Platform Presence (Search Everywhere): Perhaps the biggest shift with GEO is that Google is no longer the only game in town for organic visibility. Yes, Google Search is still huge – but a significant portion of search discovery is now happening on other platforms and will continue to diversify. AI chatbots and assistants have become new “search engines” in their own right. For instance, OpenAI’s ChatGPT (with GPT-4) can browse and retrieve information, and millions of users ask it questions daily. Bing Chat , powered by GPT-4, integrates with the web and can drive traffic when it cites sources. Perplexity AI , known as an “answer engine,” delivers responses with direct citations, and has carved out a niche of users who prefer its style of Q&A search. Other players like Anthropic’s Claude (with its 100K+ token long context) and emerging models like Meta’s Llama 2 (open-source LLMs powering various applications) mean that answers might come from a variety of AI systems, not just big search engines. Even Google itself is fragmenting: beyond the main search, we have Google’s Bard (soon to be enhanced by Gemini ), and features like YouTube’s AI summaries or Google Assistant’s evolving AI capabilities. The takeaway: you need to manage your brand presence across multiple channels and platforms**. It’s a “search everywhere” approach ( ^[12] ). Ensure that your content strategy considers not just how to rank in Google’s traditional results, but also how to appear in AI-driven answers on different platforms . This could mean optimizing content for Bing (which might have different dynamics and user base), supplying Q&A formatted content that tools like Perplexity prefer, or even providing data to voice assistants and app-based AI helpers. Marketers should track where their audience is searching – for example, younger users might be on TikTok or YouTube for certain queries, while developers might ask GitHub’s Copilot or Amazon’s Alexa. Generative AI is also entering enterprise search (e.g., Microsoft 365 Copilot for internal company info). In all cases, the fundamental strategy is to be wherever your users have questions . Don’t rely solely on one traffic source. As a recent study emphasized, winning in today’s landscape requires an omnichannel approach , capturing traffic where it’s booming (AI) while sustaining where it’s still dominant (traditional search) ( ^[15] ). Your visibility plan should span multiple engines and AI platforms so you catch your audience no matter how the search paradigm shifts. Understanding AI Models & Their Consumption of Information: A key new layer for SEO professionals is learning how AI systems ingest, interpret, and output information. Unlike traditional search algorithms, which primarily index and rank pages, large language models (LLMs) generate answers by predicting text based on patterns in their training data (and any retrieved data). This introduces challenges like hallucinations (the AI might fabricate information if it can’t find a good answer) and knowledge cut-offs (ChatGPT, for example, had limited knowledge of events after 2021 until browsing was introduced). As a marketer, you don’t need to become a machine learning engineer, but you should grasp the basics of how these models work (as covered in Chapter 4). For instance, knowing that retrieval-augmented generation (RAG) is used in some search AIs – meaning the AI will pull content from external sources in real time – tells you that keeping your content up-to-date and well-optimized for retrieval is important. It also means the context and phrasing of your content can affect whether the AI selects it. Since LLMs work heavily off patterns, if your content directly answers a common question in concise terms, it’s more likely to be picked up. On the other hand, if the content is buried in fluff or not explicitly answering the query, the AI might overlook it. Additionally, understanding that AI might rephrase or summarize your content, marketers should focus on clarity of writing and providing text that can stand on its own when extracted. Another consideration is that some AI platforms (like Perplexity, Bing, or SGE) cite sources for their answers ( ^[8] ), whereas others (like vanilla ChatGPT without browsing) do not. When citations are provided, there is a premium on being one of the sources named – it can drive traffic and brand visibility. Therefore, write content in a way that lends itself to being quotable: succinct definitions, step-by-step instructions, lists of benefits/drawbacks, etc., which an AI might choose as a concise answer to a user’s question. In summary, optimize not just for ranking algorithms, but also for AI comprehension . Think about how an AI scanning millions of words might decide your content is the best answer. This might involve using explicit question-and-answer formats, incorporating likely user questions as headings, and avoiding ambiguous language. By aligning your content format and style with AI’s way of consuming information, you add an extra layer of optimization – one that could make the difference between your content being the one an AI picks to show versus being ignored. Blending Classic SEO with New GEO Tactics: Finally, success in this new era comes from synergy – blending the best of SEO and GEO. It’s not an either/or choice. The foundational SEO practices (keyword research, on-page optimization, link earning, user-intent focus, etc.) remain highly relevant. What changes is how we execute them and where we apply them. For example, keyword research now should include researching common user prompts and questions that people might pose to AI chatbots (think of longer, natural language queries). Content optimization now means ensuring your text is structured to be extracted or summarized by an AI, not just to rank on a page. Link building and digital PR are still about building authority, but now not just to impress Google’s algorithm – it’s also to ensure your brand is well-known and trusted broadly, since AI models pick up on the reputation of sites (often indirectly through training data or citation frequency). The key takeaway is to use a hybrid strategy : continue leveraging all the “tried and true” methods of driving organic traffic, plus layer on GEO-specific strategies from this book (like prompt optimization, schema for AI, monitoring AI mention share, etc.). When done together, the whole is greater than the sum of the parts. For instance, an authoritative blog post that follows SEO best practices might rank on Google page 1 (great), but if it also is formatted in a way that a snippet can be used in an SGE overview or a voice assistant’s answer, then that single piece of content is now pulling double duty. Blending SEO and GEO ensures you’re not leaving any opportunity on the table. It’s about being thorough and forward-looking – optimizing for humans, for search engine bots, and for AI models simultaneously. Marketers who master this balancing act will create durable search strategies that flourish regardless of how the winds of technology change. In practice, this means every time you create or update content, you should be asking: Is it high-quality and relevant (for users)? Is it optimized for search indexing (for SEO)? And is it formatted/structured for easy AI consumption (for GEO)? If you can check all those boxes, you’ve achieved the SEO+GEO synergy that will carry your marketing through 2025 and beyond.

15.3 Assess and Strategize: Developing Your GEO Game Plan

With the conceptual groundwork laid, it’s time to get practical. How can you apply a GEO lens to your current strategy and identify where to go next? This section is a call to action for marketing teams: systematically evaluate how your content is performing in the AI-driven search landscape, and build a roadmap to improve your visibility. The shift to generative AI in search presents both risks (e.g. losing clicks to AI answers) and opportunities (becoming a go-to cited source in your domain). A smart strategy will maximize the opportunities and mitigate the risks. Here’s how to get started: 1. Audit Your Current Content for AI Visibility: Begin by understanding how (and if) your content is appearing in generative AI answers today. This requires a different kind of audit than a traditional SEO audit. Instead of only checking Google rankings, you’ll want to simulate queries on AI platforms . For example, try asking ChatGPT (with browsing enabled or via Bing Chat) questions related to your industry and see if your brand or content is mentioned in its answers. Do the same on Perplexity AI , which is particularly useful as it explicitly shows citations. If you have access to Google’s SGE or a lab version of their AI search, observe whether any of your site’s content is being pulled into the AI snapshots. Document these findings. You might discover, for instance, that for some important customer questions, your competitor’s blog posts are consistently getting cited by AI, while yours are nowhere to be found. This is exactly the kind of insight you need. It highlights gaps where competitors are earning trust and visibility in AI results and you aren’t ( ^[16] ). There are also emerging tools to help automate this tracking. Specialized “AI visibility” trackers can monitor when and where your brand is mentioned in AI answers across platforms (e.g., tools that track brand mentions on Perplexity, ChatGPT, Google’s Gemini, etc.) ( ^[17] ). Using such tools, or even manually spot-checking on a regular basis, will give you a baseline of where you stand. The goal of this audit is to create a map of the AI answer landscape in your niche: identify queries where you’re strong (your content is being referenced) and, importantly, queries where you’re absent but competitors are present. 2. Identify Content Gaps and Opportunities: Once you have audit data, analyze it to pinpoint patterns. For which topics or question types are competitors showing up in AI answers instead of you? Those are your content gaps and areas for improvement ( ^[18] ). Categorize these gaps: Are they high-level informational queries, or specific how-to questions? Are they early in the customer journey (e.g. “What is the best way to do X?”) or closer to conversion (e.g. “Pricing of product Y vs Z”)? Identifying the nature of the missed queries helps prioritize your efforts. For example, you might find that your site has plenty of content on broad topics, but lacks the very specific Q&A content that AI systems love to quote. Or perhaps you have good blog articles, but they’re not structured to answer concise questions (so AI ignores them). One common gap might be a lack of FAQ-style pages or how-to guides that directly answer common user questions. If you discover, say, that “How do I perform a technical SEO audit?” always returns a competitor’s checklist in AI answers, and you don’t have a comparable piece, that’s a clear content opportunity. It’s also valuable to look at the format of content that AI is choosing. Is it citing listicles, step-by-step guides, definition boxes, forums, product comparison tables? This can guide the format of content you should create. During this gap analysis, also consider emerging topics. Generative AI can handle very niche questions that users might not even type into Google traditionally. Pay attention to conversational or long-tail queries that appear in your audit – these could represent unmet needs or new trends that you can cover. In short, treat this like classic SEO gap analysis but for AI: find the queries where you aren’t the answer and make a plan to become the answer . 3. Optimize and Create Content for AI Excerpts: Armed with a list of content needs, start filling the gaps with AI-focused optimization. This doesn’t necessarily mean entirely separate content from your existing SEO content – often it’s a matter of tweaking or augmenting what you have. Here are some tactics: Embed likely prompts/questions into your content: Make sure that for each topic, your content explicitly asks and answers the common questions users have. Use headings that mirror user queries (e.g., “What is ___?” , “How to ___?” , “Best way to ___” ). This not only helps SEO (featured snippets) but also gives AI models a clear signal of relevance ( ^[19] ). Front-load the answer: Write a concise, clear answer immediately following the question. Generative models often grab the first relevant text they encounter. For instance, start a section with a direct answer in the first sentence, then provide details. This way, even if the AI only uses one or two sentences, it will likely capture the key information. Use structured formats: Wherever possible, use bullet points, numbered lists, tables, and step-by-step instructions. These formats are easy for AI to digest and present. A list of bullet-pointed benefits or steps might be directly lifted into an AI answer because it’s already in a user-friendly format. If you’re writing a “how-to,” consider a step 1, 2, 3 format. If summarizing pros and cons, perhaps use a table. These structures increase your chances of being excerpted. Incorporate Schema Markup: As noted earlier, adding structured data like FAQ schema can make your Q&A content more visible to search engines. It might also help AI find the Q&A pairs on your page. Google’s AI overviews have been observed pulling content from pages that have clear schema markup (though AI isn’t limited to that). Regardless, schema is a best practice for signaling the structure of your content to any machine reader ( ^[14] ). It’s a low-hanging fruit to implement during content creation or updates. Prioritize clarity and accuracy: AI answers need to be correct and complete. If your content leaves ambiguity, the AI might fill in the blanks (and potentially hallucinate or use someone else’s content). So ensure each piece you create or update is fact-checked and clearly written. Where appropriate, cite sources or data within your content – interestingly, if the AI picks up your content and you’ve cited a reputable source, it might treat your page as more trustworthy. Content length and depth: There’s an ongoing debate about whether long-form or short-form content is better for AI. The answer is context-dependent. Long-form content that is well-structured can cover multiple sub-questions and serve many AI queries (especially models with retrieval that can search within a page). However, if it’s too verbose without structure, the key points may be lost. So, when creating in-depth content, break it into logical sections (with descriptive headings) so that an AI can navigate to the relevant part. For targeted questions, shorter dedicated pieces (like a focused FAQ entry) might work better. It may make sense to have both: a deep pillar page and individual pages for specific questions, interlinking them (the pillar-cluster model). This way, you cover users who want detail and those (or AIs) who want just the snippet. The overarching strategy is to make your content the “ideal answer” for both people and AI. If a user asks a question, your page should be the one that a helpful assistant would naturally choose to quote because it directly and comprehensively answers that question. By auditing where you fall short and then tailoring your content to fill those gaps, you’re effectively doing GEO – Generative Engine Optimization – in practice. 4. Leverage Your Strengths – and Amplify Off-Page Signals: Optimization isn’t only on-page. Recall from earlier chapters that off-page factors (links, brand mentions, etc.) contribute to authority. That remains true in the AI era. If your website or brand is generally well-regarded online, AI models are more likely to trust and select your content. So part of your GEO game plan should be to strengthen your off-page signals in parallel. This can include digital PR campaigns to earn high-quality mentions and backlinks, engaging in communities and forums where your brand expertise can be noted, and encouraging satisfied customers to leave reviews or discuss your brand (where appropriate). AI models trained on large swaths of the internet may indirectly pick up on these signals – for example, a model might “know” that Brand X is often associated with authoritative content on a topic if it has seen Brand X referenced frequently in that context. While you can’t control an AI’s training data post hoc, you can influence the ongoing web presence of your brand. Additionally, some AI search engines like Perplexity or future ones might incorporate real-time metrics (like site authority or popularity) in choosing which sources to cite. There is already evidence that trust and source quality influence AI results (for instance, Perplexity tends to cite sources that are well-established and have relevant content across a topic). So continue to invest in off-page SEO – it’s your “brand authority” work that will pay dividends in both classic and AI search. A specific tip: monitor how your brand is mentioned in AI outputs over time. As earlier noted, tools can track this, but even setting up Google Alerts for your brand + terms like “ChatGPT” or checking social media for people sharing AI outputs can give insight. If you find your content is being cited, amplify that success – for instance, you might promote “featured in ChatGPT’s answers” if that happens (some companies have humorously done this on social media). While that might be more vanity than strategy, the point is to be aware of your progress. Conversely, if you see misinformation from AI about your brand or topic, that’s an area to address with corrective content (so that future AI have better info) or even outreach to the platform if appropriate. 5. Prioritize and Roadmap Your GEO Initiatives: Just as with any marketing plan, you need to prioritize efforts based on impact and resources. After auditing and identifying gaps, you might have a long list of to-dos – don’t get overwhelmed. Rank them. Which missing content pieces, if created, would likely yield the biggest benefit? Perhaps there’s a high-volume question where you’re absent in AI and appearing there could drive significant traffic or brand exposure. Target that first. Maybe there’s a piece of content that, with a few tweaks (adding an FAQ section or restructuring a summary), could start getting picked up by AI – that’s a quick win to implement. It could help to align your GEO roadmap with your existing content calendar or SEO projects. For example, if you were already planning a blog overhaul next quarter, integrate the GEO-oriented changes into that project. Set specific goals, such as “By next quarter, improve our representation in AI answers for 10 key industry questions” and track progress. Remember to use an experimental mindset: you might update some content specifically to be more AI-friendly, and then test by querying the AI platforms again to see if it now gets picked up. GEO is new for everyone, so there will be some trial and error – and that’s okay. The important part is to start now . Use the outline of this eBook and the chapter checklists as a starting framework for a GEO action plan. If your team is large, consider forming a small “AI search SWAT team” or assigning an “AI search champion” who keeps abreast of the latest developments (since features like SGE or Bing’s AI are evolving rapidly) and coordinates the GEO efforts. Conduct training sessions to get your content writers and SEO specialists up to speed on GEO concepts, so everyone is thinking about AI visibility when creating content. Treat this as a new dimension of your SEO program – much like how years ago social media or mobile optimization became part of the job. By formalizing it into your strategy, you ensure it gets the attention it warrants. 6. Continuously Monitor, Measure, and Refine: Optimization is never “one and done,” and GEO is no exception. After implementing changes, keep monitoring how your content performs in both traditional and AI search. Did your traffic from organic search change after SGE rolled out? Are you seeing referral traffic from new sources (like Bing Chat or Perplexity)? Some analytics tools may not yet label these clearly, so you might have to dig – for instance, Bing Chat traffic might show as referral from Bing or with certain user-agents, and Perplexity traffic might appear as direct or referral with a specific pattern. Stay informed about measurement techniques (Chapter 13 discussed emerging ways to track AI visibility). You could establish some KPIs, such as “number of times our brand is cited in AI results per month” (if trackable), or simply the organic traffic from search vs. AI-driven sources. Also measure engagement: if AI is sending fewer clicks, are those clicks more qualified? (One study found generative AI traffic, while smaller in volume, converted 23% better than organic search traffic ( ^[20] ) – indicating high intent users.) That might influence your content focus. Importantly, be prepared to iterate. If a certain piece still isn’t getting traction in AI, analyze why – maybe the competitor’s content is genuinely more in-depth or the AI has a preference you haven’t matched. It might take a couple rounds of refinement. On the flip side, if you suddenly gain a top spot in an AI answer, capitalize on it – ensure the content is updated frequently so it stays relevant, and maybe link from that page to other related pages (to pull up additional content with it). And of course, as new AI features roll out (say, Google enabling follow-up questions with sources, or new entrants like Neeva AI – oh wait, Neeva shut down, but others will come!), you will need to adjust strategy. Build into your roadmap periodic GEO strategy reviews , maybe quarterly, to assess what’s changed in the landscape and how you need to pivot. By going through these steps – auditing, identifying gaps, optimizing content, bolstering authority, planning and prioritizing, and continually measuring – you can develop a comprehensive GEO game plan tailored to your organization. This proactive approach ensures that as generative AI becomes more ingrained in search behavior, your company will not only keep up but lead. The aim is to have your content be the one that the AI picks – and that won’t happen by accident. It requires strategy and effort, but the payoff is maintaining your visibility and traffic in a world where search results are increasingly AI-curated. Think of it this way: We spent decades optimizing for Google’s algorithm. Now we have a multitude of “algorithms” (LLMs and AI systems) to consider. It’s a broader playing field, but also one with more opportunities for those willing to play. Use this book’s insights as a roadmap, get your team on board, and start implementing GEO tactics step by step. In doing so, you’ll transform what could be seen as a threat – AI taking clicks – into an opportunity: AI driving engaged users to your content, and even recommending your brand in contexts where traditional search might not.

15.4 Continual Learning and Adaptation

If there is one truth in digital marketing, it’s that nothing stands still for long . The rise of generative AI in search over the past two years is a case in point – few could have predicted at the start of this decade that we’d soon be interacting with search engines via chat interfaces that write answers like a human. Yet here we are, and this rapid pace of change is only set to continue. For search marketers, the implication is clear: a mindset of continual learning and experimentation is not just valuable, it’s essential. This chapter (and this eBook) provides a snapshot of the landscape in 2024–2025, but new developments are certainly around the corner. How can you and your team stay ahead of the curve? First, stay curious and informed . Make it part of your routine to follow the latest news on search engine updates and AI advancements. Subscribe to industry newsletters (e.g., Search Engine Land’s updates, SEO Roundtable, etc.), listen to marketing and AI podcasts, attend webinars or conferences (many now have dedicated AI and SEO tracks). For example, Google’s AI search updates often roll out with announcements – being aware of a change like “SGE now rolling out to all users” or “New citation feature added to Bard” can alert you to adjust your strategy. Similarly, keep an eye on OpenAI, Microsoft, and other big players’ announcements; when ChatGPT gets a new capability (like the addition of Plugins or multimodal input), ask how that could affect search behavior or content consumption. In late 2023, for instance, OpenAI introduced web browsing and plugins for ChatGPT, suddenly enabling it to pull live information and interact with websites for answers. Marketers who quickly experimented with those features learned how their content might be accessed or what kind of queries were now possible (e.g., users asking ChatGPT to compare live product prices). By staying informed, you position yourself to respond rapidly to changes, rather than playing catch-up. Second, embrace a culture of testing and learning . The beautiful (and sometimes frustrating) thing about working with AI is that it often requires hands-on experimentation to truly understand. Encourage your team to play with these tools regularly. For example, you might have monthly “lab days” where the team does nothing but experiment with different prompts on ChatGPT or tests how a new search feature works. Document the results and share internally. Ask questions like: How does ChatGPT’s answer differ when I supply a source versus when I don’t? Does Bing Chat cite my site if I phrase a query one way but not another? What kind of content does Google’s AI Overview refuse to show (perhaps because it’s YMYL – “Your Money or Your Life” content that needs high authority)? This kind of exploratory testing can reveal quirks and opportunities. It’s analogous to how SEO professionals conduct A/B tests for titles or content – now you might be testing prompts or content structures for AI. The key is to treat the AI platforms as new arenas for optimization and to continuously poke at them to discover how they tick. A practical example of learning through experimentation: Suppose you notice that when you ask Perplexity AI a question about a topic your site covers, it cites Wikipedia and a competitor, but not you. Try tweaking your content and see if that changes after Perplexity reindexes (Perplexity tends to update relatively quickly since it’s pulling from the live web for many queries). If after adding an FAQ section or making your answer more concise, you then see Perplexity citing you, voila – you’ve learned something actionable and improved your presence. The same goes for Google’s SGE: maybe you find that adding structured data got your image or snippet included in the AI result. These little wins come from that experimental, persistent approach. Don’t be afraid to get your hands dirty with these tools – think of it as “prompt SEO” or “AI UX testing.” The insights you gain will inform your broader strategy and keep you agile. Additionally, foster knowledge sharing and continuous education within your organization. This might involve upskilling team members. For instance, your SEO team might need to learn some basics of prompt engineering or get comfortable interpreting AI outputs. Conversely, your content writers might benefit from understanding how language models work so they can write in an AI-friendly manner. Encourage cross-functional learning: maybe your data science team (if you have one) can brief the marketers on how LLMs are being used internally or in the market. Create a space for sharing articles or case studies about AI in search – perhaps a Slack channel or weekly email where anyone can drop interesting tidbits they’ve found (like “Hey, check out how Expedia’s plugin on ChatGPT works – could this be something we consider?”). By making learning a continuous, communal process, you ensure that you’re not relying on one “AI expert” who might overlook something; instead, you have a team of savvy marketers all keeping eyes on the horizon. It’s also wise to maintain a healthy dose of skepticism and critical thinking amidst the AI hype. Not every trend will pan out, and not every shiny AI tool will yield ROI for your business. Part of continual learning is discerning signal from noise. For example, many AI tools and platforms will emerge claiming to be “the next big search disruptor.” It’s okay to try them, but use data to decide if they warrant significant effort. You might remember how voice search was hyped to extremes a few years back – “By 2020, 50% of searches will be voice,” was a claim that led some to obsess over Alexa optimization. The reality turned out differently. Similarly, as you experiment, keep asking: Do we see actual user adoption of this? Is this likely to impact our target audience? If an AI search app has a tiny user base and low growth, you might not invest heavily there. On the other hand, if you see something gaining traction (like how ChatGPT went from zero to 100 million users in two months – an unprecedented adoption rate), you know it’s time to pay attention. Use the metrics and studies available: for instance, earlier we cited how generative AI traffic is still under 1% of total traffic on average ( ^[21] ), but growing extremely fast. That suggests it’s currently a small slice, but one that’s increasing – a classic “emerging channel” scenario. Thus you allocate resources in proportion: not abandoning the old (since 99% of traffic might still be from traditional sources) but preparing for the new (since that <1% could become 5%, 10%, etc., and you want to be ahead of competitors in capturing it). Continual learning helps you make these judgment calls with better context. Another aspect of adaptation is being ready for organizational change . Just as companies eventually built mobile-specific teams or roles (like a “Head of Mobile Strategy”) once mobile became big, we may see the rise of roles like “Head of AI Search Optimization” or “LLM Content Strategist”. Smaller organizations might not have separate roles, but people will wear multiple hats. Be open to shifting job scopes and workflows. Your SEO specialists might need to coordinate more with developers (for implementing schema or API-based content delivery for AI), or with PR teams (for boosting brand authority signals). Breaking silos between SEO, content, PR, and data will become increasingly important, because GEO spans all those areas (technical, content, off-page, analytics). Encourage your team members to acquire interdisciplinary skills – e.g., an SEO who can also write effectively for FAQs, or a content writer who knows how to do basic schema markup. Teams that adapt skill-wise will execute faster on GEO tactics than those stuck in rigid role definitions. Finally, keep an optimistic and proactive attitude . It’s easy to feel intimidated by the speed of AI progress – even professionals feel like it’s hard to keep up. But remember, you are not alone in figuring this out; the entire industry is learning together. By maintaining a learner’s mindset, you actually make your job more exciting – there are new discoveries to be made, and you can innovate in ways that weren’t possible before. Celebrate small victories in your adaptation journey. Did someone on the team get our content featured in a Bing chat result? Share it and acknowledge that win. Did traffic from an AI assistant convert a big lead? Highlight that success. These morale boosters reinforce the value of what you’re doing and keep the team engaged in the process of continuous improvement. In summary, the only constant in search (and in marketing tech) is change. Preparing for the future means never feeling you’ve “arrived” at a perfect strategy – there is always something new around the corner. By staying informed, testing boldly, learning as a team, and remaining adaptable, you build a kind of resilience. No matter what the AI companies or search engines throw your way next, you’ll be able to pivot, experiment, and find a path to success. Today it might be SGE and ChatGPT, tomorrow it might be something like personalized AI agents or new regulations (more on those next) – in all cases, your best asset will be a culture of continuous learning . In the fast-moving AI landscape, today’s cutting edge can become tomorrow’s old news . But with an agile mindset, you’ll ride each wave and keep on leading rather than lagging.

15.5 Looking Forward with Confidence

As we conclude this guide, it’s worth casting our eyes to the horizon. What does the future hold for search and how can marketers position themselves to thrive? While no crystal ball is perfect, we can anticipate several trends and frontiers that are likely to shape the coming years. The encouraging news is that by focusing on providing real value and staying attuned to how people and machines interact, online marketers can navigate whatever comes next. The tools, interfaces, and acronyms may change, but the core goal remains timeless: connect your audience with the information and solutions they seek , wherever and however they search. One major theme is the rise of personal AI assistants and private agents . We’re already seeing early signs: OpenAI’s vision is to eventually provide personal AI that can act as a knowledgeable companion. Apple is rumored to be working on more advanced AI integrations for Siri, and numerous startups are creating AI assistants tailored to specific domains (health, finance, etc.) or even individual personalized AI that learns a single user’s preferences. In the enterprise space, tools like Microsoft 365 Copilot act as an AI assistant that knows your company’s data. As these personal and private AI agents become more common, the notion of “search” might shift from typing into a search bar to simply asking your own device or software to find or perform something. Imagine a future where a business owner asks their AI assistant, “Give me the top three recommendations from Gartner for CRM systems and schedule demos,” or a consumer asks, “Find me a vacation package in my budget and book it if it looks good.” The AI might do all that without the user ever visiting a traditional website. What does this mean for us? It means our content and services will need to be accessible to these AI intermediaries. This could involve partnerships and integrations (for example, ensuring your e-commerce site can interface with voice assistants or offering APIs that personal agents can tap into). It also underscores even more the importance of structured data and feeds – if you want a personal AI to recommend your product, you might be providing a data feed to some central repository, or ensuring your reviews and specs are well-formatted for AI consumption. It’s a continuation of GEO: you’re optimizing not just for public search engines, but potentially for individual AI agents that pick and choose content for their single user. While it sounds far-out, signs point to this being a logical extension of current trends (consider how some people already rely on Alexa or Google Assistant heavily – amplify that by 100× in intelligence). Marketers should keep an eye on this and perhaps pilot integrations with any emerging platforms for personal assistants. Another trend is multimodal search and visual answers . Large models are getting better at handling images, audio, and video in addition to text. We’ve seen Google incorporate Google Lens (image recognition) into search results, and Bing’s chat can understand images (you can literally show it a photo and ask questions). Future search experiences might seamlessly blend text, images, and video. For instance, a user might ask an AI search, “How do I fix this?” while showing a photo of a broken appliance, and get an answer with a step-by-step explanation alongside a diagram or video clip. Already, tools like Bing can generate images (via DALL-E) as part of answers, and Google has demonstrated showing visual elements in SGE answers (like product images, maps for local queries, etc.). For marketers, this means two things: visual content SEO and optimization for multimodal contexts . Ensure your images are optimized (good alt text, descriptive file names, schema for images if relevant) because AI might pull an image from your site to include in an answer. Same for video – maybe the AI will quote from a transcript of a YouTube video or show a snippet. If you produce video or audio content, making transcripts and marking them up could increase your content’s accessibility to AI. Also, consider creating visuals that are informative (infographics, charts) since an AI might choose to present a user with a relevant graphic if available. The line between “search result” and “media content” is blurring. A forward-looking content strategy might involve a blend of text and visuals designed to answer questions. Imagine a knowledge base article that contains a summary, a quick infographic, and a short video – a comprehensive answer package. Such content might be ideal for an AI to draw upon, because it has multiple ways to satisfy a user’s query (some users prefer text, some visual). We can foresee AI search tools evolving to use whatever medium best answers the question at hand. Voice search, which we touched on, might see a renaissance (Voice 2.0) thanks to AI. Early voice assistants were often frustrating, handling only simple queries. But with LLMs, conversational voice interaction is far more feasible. ChatGPT’s style of dialogue could translate to a voice assistant that actually has a back-and-forth conversation. This means voice SEO could come back into focus – not in the old sense of “position zero” answers only, but in ensuring your content can be read aloud in a satisfying way. If someone using voice asks, “What’s the best SUV for a family of four?” the AI might narrate a short comparative answer (sourced from web content). Content that is written in a natural, conversational tone might perform better for voice answers (since it’s literally going to be spoken). Additionally, local search via voice (e.g., in cars or smart speakers) might get more integrated with generative AI, making results more nuanced. For example, instead of just getting the nearest pizza shop, a voice AI might advise, “The nearest pizzeria is Mario’s Pizza, which has 4.5 stars. By the way, they have a special on Tuesdays. Would you like to hear the menu or make a reservation?” That kind of interaction would draw on various data sources – your Google Business Profile info, online reviews, your website’s menu – all summarized by AI. Businesses should prepare for such scenarios by maintaining consistent, machine-readable info online (good local SEO, rich data) and perhaps even integrating with voice platforms. Another critical aspect of the future is regulation, transparency, and ethical AI practices . Already, governments are scrutinizing AI’s impact on privacy, competition, and content rights. The EU’s AI Act and initiatives by regulators like the UK’s CMA (Competition and Markets Authority) are looking at how AI-generated search summaries use content and whether that’s fair to publishers ( ^[22] ) ( ^[23] ). Publishers and companies like OpenAI have tussled over content usage (e.g., Reddit and Stack Exchange adjusting API access because AI companies were training on their data). We can expect some form of guidelines or requirements to emerge around transparency – for instance, AI-generated search results may need to clearly cite sources or even seek licenses for certain content. Google has already added source links in SGE, and other tools like Perplexity made citations a core feature. Marketers should advocate for and welcome transparency , because being cited is how we get traffic and credit. It’s possible that future AI search tools will offer publishers options, like opting out (via something akin to robots.txt for AI) or opting in for enhanced citation. Keep an ear to the ground on these policy developments. Also, consider the ethics and user trust: users might begin to demand to know why an AI recommended something. Ensuring that your content is honest, well-sourced, and trustworthy isn’t just good for SEO – it could be vital for passing whatever quality thresholds are set in a more regulated future. If, say, an AI is obliged to only show content from sites that meet certain criteria (to avoid misinformation), you want your site to be on that safe list. That likely comes down to continuing what Google has long advised: make content that is helpful, accurate, and people-first ( ^[24] ). Hand in hand with regulation is the push for better attribution and possibly compensation . There are ongoing discussions about how publishers might be compensated if AI truly starts drawing large amounts of value from their content without clicks. We may see search engines exploring models like traffic sharing (already, SGE when it cites sources encourages clicks by showing the source links – it’s up to us to create compelling titles and content that make the user click through). Or perhaps new partnerships will form (like how Google licenses news content in some countries for Google News). While the individual marketer can’t control this, being aware and agile means you can adjust if, for example, a major platform changes how it displays content or offers an affiliate-like program for content used in AI answers. Finally, amid all this tech, the human element – creativity, empathy, and strategy – remains your ace. AI is great at scaling information and even generating passable content, but it often lacks the true creativity and insight that a human can bring. Your competitive edge as a marketer will increasingly be the creative strategies and original ideas that differentiate your brand. Use AI as an augmenter: let it handle rote tasks or give you quick drafts, but infuse your campaigns with human creativity that AI cannot easily replicate. That might mean unique brand storytelling, building community and loyalty (which no AI can steal – if users love your brand, they might specifically seek it out), or leveraging emotional intelligence in your copy and visuals. Chapters 9 and 12 discussed focusing on content AI can’t easily copy and doing things ethically – those principles guard your long-term differentiation. For instance, an AI can aggregate facts about all running shoes and even generate a generic article “Top 10 Running Shoes in 2025”. But only you (and your team) can produce genuinely novel research, or have an actual opinion piece interviewing athletes, or create a fun interactive quiz that engages users – those kinds of assets make your site stand out such that even if AI summarizes the topic, users will seek the depth or personality you provide. In the future, as more commodity content gets handled by AI, truly exceptional content (whether due to creativity, depth, or exclusivity) becomes even more valuable. It’s like how stock photos are abundant, but a real custom illustration or a viral video has much more impact – quantity vs quality. Therefore, looking forward with confidence means betting on human strengths complemented by AI capabilities. The search and AI landscape will continue to change – perhaps faster than ever – but if you build a foundation of adaptability, continuous improvement, and user-centric strategy, you will navigate it successfully. To recap the mindset: embrace change, keep learning, stay ethical, and focus on real value . By doing so, you ensure that no matter what tools people use to search – be it a keyboard, voice, or mind-reading AI (who knows!) – your efforts are aligned with the fundamental goal of meeting user needs. The channels will shift, but that mission does not. In conclusion, the journey from SEO to GEO is an ongoing one. This chapter – and this eBook – is not an end but a beginning. Now armed with an understanding of how search is evolving and concrete strategies to adapt, you have the roadmap to move forward . Embrace the synergy of old and new, lead your team in strategic innovation, and maintain the spirit of serving your audience. By doing so, you can look to the future not with fear, but with optimism. The era of generative AI in search holds immense possibilities for those ready to seize them. As we move ahead, remember that search marketing has always been about connection . Generative AI is just another way people seek connection – connection to information, to services, to answers. Your job remains to forge that connection. Do that, and you will not only survive the next frontiers of search – you will flourish.

References

^[1] WebFX - WebFX.com URL: https://www.webfx.com/blog/seo/gen-ai-search-trends/

^[2] BrightEdge - BrightEdge.com URL: https://www.brightedge.com/news/press-releases/new-report-brightedge-reveals-surge-ai-search-engines-signaling-new-era-online

^[3] IBM - IBM CEO Ginni Rometty URL: https://www.ibm.com/thought-leadership/institute-business-value/c-suite-study/ceo

^[4] SparkToro - SparkToro.com URL: https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/

^[5] Adweek - Adweek.com URL: https://www.adweek.com/performance-marketing/google-zero-click-2025-seo/

^[6] Google Search Central Blog - Google.com URL: https://blog.google/products/search/generative-ai-google-search-may-2024/

^[7] BrightEdge - BrightEdge.com URL: https://www.brightedge.com/blog/10-observations-about-transition-sge-ai-overviews-may-2024

^[8] Ahrefs - Ahrefs.com URL: https://ahrefs.com/blog/google-ai-overviews/

^[9] SE Ranking - SeRanking.com URL: https://seranking.com/blog/ai-overviews/

^[10] Statista - Statista.com URL: https://www.statista.com/statistics/1454204/united-states-generative-ai-primary-usage-online-search/

^[11] Google Search Central Blog - Google.com URL: https://blog.google/products/search/generative-ai-search/

^[12] Single Grain - SingleGrain.com URL: https://www.singlegrain.com/search-everywhere-optimization/search-everywhere-optimization-tactics-for-growth-in-2025/

^[13] Search Engine Land - SearchEngineLand.com URL: https://searchengineland.com/evolving-seo-for-2025-what-needs-to-change-450911

^[14] Schema App - SchemaApp.com URL: https://www.schemaapp.com/schema-markup/the-semantic-value-of-schema-markup-in-2025/

^[15] Single Grain - SingleGrain.com URL: https://www.singlegrain.com/digital-marketing-strategy/answer-everywhere-optimization-the-complete-framework-for-digital-visibility/

^[16] BeebyClarkMeyler - BeebyClarkMeyler.com URL: https://www.beebyclarkmeyler.com/what-we-think/guide-to-content-optimzation-for-ai-search

^[17] KI Company - Ki-Company.ai URL: https://www.ki-company.ai/en/blog-beitraege/schema-markup-for-geo-optimization-how-to-make-your-content-visible-to-ai-search-engines

^[18] Edge45 - Edge45.co.uk URL: https://edge45.co.uk/insights/optimising-for-ai-overviews-using-schema-mark-up/

^[19] GetPassionFruit - GetPassionFruit.com URL: https://www.getpassionfruit.com/blog/faq-schema-for-ai-answers

^[20] WebFX - WebFX.com URL: https://www.webfx.com/blog/seo/gen-ai-search-trends/

^[21] Adobe Analytics - Adobe.com URL: https://business.adobe.com/blog/the-explosive-rise-of-generative-ai-referral-traffic

^[22] European Commission - Digital-Strategy.ec.europa.eu URL: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai

^[23] Osborne Clarke - OsborneClarke.com URL: https://www.osborneclarke.com/insights/regulatory-outlook-may-2025-artificial-intelligence

^[24] Google Search Central - Developers.Google.com URL: https://developers.google.com/search/docs/fundamentals/creating-helpful-content