One recent afternoon, a marketing director at a mid-sized software firm decided to ask ChatGPT the kind of question she used to type into Google: “What’s the best project management tool for a growing business?” The AI chatbot responded with a confident, one-paragraph recommendation of a well-known app. Conspicuously absent was any mention of her own company’s software – a product that had long ranked on Google’s first page for similar queries. In that moment, she glimpsed a new reality of the internet: in the age of AI-driven search, a business can be invisible not only on the second page of results, but completely invisible if it’s not part of a chatbot’s single answer. 

For decades, search engines have been the primary gateway for anyone seeking brands, products or services online. Companies large and small poured resources into search engine optimization (SEO) to secure a coveted spot among Google’s ten blue links. Not long ago, a top-three Google ranking could capture anywhere from 18% to 40% of clicks for a given query. But user behavior is starting to shift. Instead of scrolling through pages of links, many people are now turning to artificial intelligence assistants that deliver one synthesized response. In 2025, some 71.5% of people surveyed said they have used AI tools like ChatGPT, Claude or Perplexity for search – especially for quickly comparing products or getting recommendations. 

The most popular of these, OpenAI’s ChatGPT, has seen its web traffic roughly double over the past year, reaching nearly 6 billion visits in September 2025. And while Google still dominates overall search volume, it’s racing to incorporate AI itself: the company’s new Search Generative Experience now injects AI summaries into many queries, and a highly anticipated chatbot known as Gemini is on the way. Analysts predict these trends will only accelerate. Gartner, a research firm, forecasts that traditional search could lose about 50% of its share to AI-driven tools by 2028. Even Apple has reportedly considered adopting an AI-powered search engine as the default on iPhones, in place of Google.

For users, interacting with a chatbot can be a timesaver – why wade through a dozen links when an AI can distill the web’s knowledge into a single, conversational reply? But for businesses, this “one answer” dynamic poses a stark challenge. If your brand isn’t part of that answer, it might as well not exist in that moment. As Nick Taylor of Edelman noted in a recent report, AI tools “don’t return pages of links” like a traditional search engine; instead they offer concise, curated replies that read like definitive recommendations. In other words, the AI isn’t just aggregating information – it’s effectively picking winners and losers. And if your company isn’t the winner, you risk being invisible to the consumer asking the question.

This represents a fundamental break from the old search paradigm. On Google or Bing, even if you weren’t the top result, you still had a fighting chance to catch a curious click as one of many links on page one. A generative AI, by contrast, might mention only one or two brands – or none at all. In fact, a recent analysis by BrightEdge found that ChatGPT’s answers mention an average of just 2.4 brand names per query, whereas Google’s AI-infused search overview lists about 6 brands on average. Often ChatGPT will give advice without naming any specific product or company (it was “silent” on brand names in 43% of the questions studied). Those dozens of competitors that would have appeared as search results are essentially filtered out of the conversation. The implications for visibility are enormous: it’s a winner-take-all (or winner-take-most) scenario.

Being one of the few brands that does get featured by an AI engine can have tangible benefits. When Google’s search AI (in the form of an “AI Overview” box) includes a citation or mention of a brand, that brand tends to reap more clicks if users still seek more information. A recent study showed that companies cited in Google’s AI answers enjoyed 35% more organic clicks than those left out. But those not cited saw their traffic plummet. Overall, the presence of an AI answer dramatically reduces the need to click any link: Seer Interactive, a marketing agency, found that Google results with AI summaries saw organic click-through rates drop by 61%, and even paid ad clicks dropped 68%. Strikingly, even searches without an AI box showed a 41% decline in clicks year-over-year, a trend attributed to users getting answers from ChatGPT or other platforms before they ever type a query into Google. In short, a large slice of web traffic that businesses used to count on is simply vanishing – siphoned away by answer engines that resolve queries without the need to visit a website.

This shift is already producing winners and losers, sometimes in unexpected ways. Jan van der Crabben learned this the hard way last year. He runs World History Encyclopedia, a nonprofit education website that had long depended on Google search traffic. When Google began rolling out its AI-generated summaries, van der Crabben’s site was initially thrilled to be featured as an information source. That thrill turned to alarm when he saw the numbers: in November, the site’s Google traffic dropped 25%. The AI was answering history questions by synthesizing content from his and other sites – and far fewer readers were clicking through to the encyclopedia itself. “There used to be this implicit agreement between publishers and Google” that Google could index and excerpt content and “in return, they would send traffic,” van der Crabben said. “Now, this unspoken contract is kind of breaking”. In his case, appearing in an AI answer didn’t translate to visits; it meant the AI gave away the information, eroding the incentive for users to read more on his site.

World History Encyclopedia’s story is an early warning for all organizations that rely on organic search visibility – not just media publishers. A software-as-a-service startup or a consulting firm might similarly find that a chatbot is dispensing advice or answers drawn from its website (or from its competitors’ sites) with zero attribution. A legal clinic that once earned clients by ranking atop Google results for “best estate lawyer in Denver” could be overlooked entirely if a user asks an AI assistant the same question and gets an authoritative-sounding recommendation for a different firm. In these scenarios, businesses don’t even know they were in the consideration set, because the user never saw them. The discovery process has been reduced to a single step: the AI’s recommendation.

Concerningly, companies may not realize they’ve become invisible in these channels because the change is happening behind the scenes. Unlike Google search rankings – which are public for anyone to see – the inner workings of an AI model are often a black box. If an AI chatbot “decides” your product isn’t worth mentioning, there’s no easy way to know, unless you actively go looking for how these models talk about you. And if the model is basing its answer on outdated or incorrect information, your reputation could suffer without you even hearing the falsehood. (Generative AI systems have been known to confidently assert inaccuracies – a phenomenon known as hallucination – and if those involve your brand, the misinformation could stick in consumers’ minds.) The conversational, authoritative tone of AI answers can make such statements feel definitive. All of this raises the stakes for brands: much of this shift is occurring outside the traditional analytics that marketing teams monitor. In effect, a chunk of your audience might have already moved to an AI discovery model, and you wouldn’t see it in your web traffic reports.
To be sure, the traditional search engine isn’t dead yet. Far from it – Google still handles an estimated  searches per day, dwarfing the volume of queries posed to ChatGPT. Many ChatGPT users also continue to “google” things; one survey found that 95% of people who use ChatGPT still rely on Google as well. And importantly, AI chatbots so far generate relatively few direct clicks or customers for businesses.

According to one digital marketing analysis, AI-based search referrals accounted for less than 1% of overall website traffic for major publishers, and those visits produced virtually no conversions (e.g. purchases or sign-ups). In other words, at this stage people often treat chatbots as a research tool – they might ask ChatGPT or Bing Chat for a summary or idea, but then still go to a website when it’s time to buy something or get details. All of this suggests that we are in a transitional period: the old search habits are coexisting with new AI-driven habits. Yet the trajectory is clear enough that companies dependent on online visibility feel they ignore it at their peril. As van der Crabben puts it, search isn’t broken, but it’s “being rewritten”, and everyone from one-person blogs to Fortune 500 brands will need to adapt.

Facing this upheaval, businesses and marketers are beginning to experiment with a new playbook for the AI era. It goes by a nascent term: Generative Engine Optimization, or GEO. Much like SEO emerged 20 years ago to help companies climb Google’s rankings, GEO is emerging now as a set of strategies to ensure companies show up in AI-generated answers. “This practice helps [brands] transform their presence inside LLMs (large language models) and… ensure the right messages are surfaced prominently” in tools like ChatGPT, Perplexity and Gemini, according to Edelman’s AI research team. In plain terms, the goal is to make sure that when someone asks an open-ended question to an AI, the answer they get includes your brand (and portrays it accurately and favorably).
How does one influence an algorithm that has no visible “results page” or ranking factors? Early GEO efforts focus on two broad fronts: information sources and technical signals. First, companies are trying to seed the right sources. Generative AI models don’t have their own human-like judgment – they rely on patterns in data. So an AI’s recommendations of, say, “the best CRM software” will be drawn from what it has ingested: news articles, reviews, forum discussions, product descriptions and so on. Studies show that community and review platforms have an outsized impact. One analysis of 680 million AI-generated answers found that platforms like Reddit and consumer review sites (for example, G2) appeared frequently as cited sources in ChatGPT, Perplexity and Google’s AI results. These are places where the AI “learns” which brands are talked about and in what context. The takeaway for businesses is that authority beyond your own website is crucial. If your company is being recommended and discussed on trusted forums, in trade publications, and by influencers or satisfied customers on review sites, an AI is far more likely to recognize and recommend it. By contrast, if your brand has a minimal footprint outside of your own webpages, the AI may have no basis to include you in an answer about your industry.

In practical terms, this means marketers are investing in what used to be called earned media – coverage and mentions that can’t be bought outright. For instance, encouraging happy customers to leave detailed reviews (with specifics about your product) on sites like G2 or Capterra can make a difference, since those details might be picked up in an AI’s training data or real-time search. Participating in relevant Reddit or StackExchange discussions (transparently and helpfully, not just self-promotion) can plant your expertise in the communal knowledge pool. Traditional PR is making a comeback as well: getting your company featured or quoted in major media or niche industry blogs creates authoritative signals that algorithms notice. All of this widens the trail of digital breadcrumbs connecting your brand to key topics, which in turn increases the odds an AI will cite or mention your brand when those topics come up.

Then there’s the technical side of GEO – making your own content more legible to AI systems. In many ways, these practices resemble classic SEO, with a new twist. Companies are ensuring their websites clearly state who they are and what they offer in plain language, particularly in page titles and headings that algorithms give weight to. If you’re a cybersecurity firm for banks, for example, your homepage should explicitly say something like “Cybersecurity Solutions for Financial Institutions” rather than just “Welcome to [Company Name]” – both people and AI crawlers need unambiguous context. Additionally, webmasters are looking at more structured ways to feed information to AI. One approach is adding Schema.org structured data markup to pages – essentially a layer of metadata that spells out facts (like “Organization = X, Product = Y, CEO = Z”) in a standardized way. This markup was originally designed to help search engines and voice assistants, but it may prove useful for AI models as well. Microsoft’s Bing, for instance, has indicated that schema markup helps its AI “understand content” on websites. And Google’s next-generation model Gemini is reported to draw on Google’s Knowledge Graph – a giant database of facts that is partially built from structured data on the web. In other words, if you provide machine-readable facts about your business (using schema tags for your products, reviews, location, and so on), you increase the likelihood that AI systems will correctly incorporate those facts into answers. This is still an emerging tactic, and some in the industry debate its current impact, but many companies see it as future-proofing their content for an AI-centric web.

What you cannot do, unlike in the old search world, is simply pay to be visible. There is no equivalent of buying a Google ad that will guarantee your brand is mentioned by ChatGPT or Claude. As Edelman’s report pointed out, visibility in LLM answers “can’t be bought” – as much as 90% of the content driving an AI’s outputs comes from organic, earned sourcesn. This levels the playing field in some respects (deep-pocketed advertisers don’t automatically win), but it also meas companies have to compete on content and credibility. Interestingly, some large publishers have begun negotiating licensing deals with AI platform providers to ensure their content is included and credited. Those deals sometimes promise a degree of prominence or attribution in AI results, which could hint at a future where certain brands secure preferential treatment. But for most businesses, the more viable strategy is to earn that prominence by being genuinely useful and present in the places AI looks.

There is also a defensive aspect to GEO. Brands are starting to monitor how AI platforms portray them, akin to how they monitor social media or online reviews. This might involve running periodic queries on popular chatbots (“What’s the best budget smartphone?” or “How does [Our Company] compare to [Competitor]?”) to see what comes up. New tools and services are springing up that promise to track a brand’s “share of voice” in AI search results, highlighting gaps or misconceptions. The first step to optimizing is knowing where you stand – some companies are surprised to find that an AI has picked up an obscure forum complaint or an outdated description from years ago. By auditing AI outputs, they can then devise content updates or PR responses to correct the record. In essence, GEO isn’t just about getting into the AI answers; it’s about making sure the right information gets in as well.

Crucially, the playbook is still being written. “The rules of the game are being written (and will be re-written) as AI continues to advance and influence habitual behaviors,” the Edelman team noted. In the early going, that means experimentation. Some forward-thinking brands are already running trials – for example, publishing Q&A-style content on their sites to see if it gets picked up by chatbots, or optimizing their presence on Wikipedia and other data sources that feed AI models. Those who move early may gain a competitive edge. Just as the first companies to embrace SEO or social media marketing often reaped outsized benefits, there’s a “huge whitespace opportunity” now for companies willing to invest in AI visibility before it’s mainstream. Early adopters can establish strong AI reputations and customer trust, while laggards risk playing catch-up in a few years when generative search may be ubiquitous.

Despite the rapid rise of generative AI search, it’s worth noting that we are essentially living through a grand experiment in information delivery. Many unknowns remain. Will users ultimately prefer a single AI-synthesized answer to the freedom of browsing multiple sources? So far, surveys suggest people like the convenience – in one poll, 72% of searchers said they use Google’s AI summary when it’s available. And a (perhaps optimistic) slice of users, over 60%, told one researcher they use an AI chatbot every day. But it’s also true that trust in AI is not absolute. Concerns over accuracy and bias mean some users still verify chatbot answers against other sources. In high-stakes domains – health, finance, legal – many people want that second opinion from a known authority. This implies that the trajectory of AI search might vary by context: it could dominate for casual, quick queries (“how to boil an egg” or “best wireless headphones under $100”), while traditional search or specialist websites remain important for more complex or sensitive inquiries.

There is also the question of transparency. Unlike a search engine results page, where sources are listed, a raw AI answer can feel like it came from thin air. Under public and regulatory pressure, AI providers may move toward more explicit citation of sources. We’re already seeing differences in approach: Google’s AI Search Overview often includes footnoted sources for facts or statements, sometimes listing a dozen or more citations per answer. ChatGPT, on the other hand, typically provides answers with no direct citations unless specifically asked. If the industry gravitates toward more citations in AI answers, that could somewhat lessen the “vanishing web” effect – users might regain pathways to click out to source websites. Indeed, in that BrightEdge study, Google’s AI gave far more citations (on average) than brand mentions, whereas ChatGPT provided far more uncredited statementsbrightedge.com. A more transparent AI ecosystem would reward companies that publish high-quality, factual content on their own sites, since those would be the sources linked in an AI summary. It might also spur new forms of partnership between content creators and AI platforms (as in those early licensing deals).

On the flip side, if AI answers become the norm and users grow comfortable taking them at face value, businesses may have to adapt to a world of drastically lower traffic and visibility. This could push more companies to shift their models – for example, toward membership or subscription services, or offline brand-building – to compensate for lost organic reach, much as World History Encyclopedia is exploring alternative strategies beyond search. In an extreme scenario, one can imagine the “open web” of many interconnected sites giving way to a kind of AI-mediated interface, where the AI becomes the primary intermediary between consumers and content. That possibility has sparked debate and even dread among internet veterans. But it’s also spurring innovation: new standards (perhaps meta-tags for AI or content watermarks) could emerge to ensure that AI systems act more like collaborators to websites rather than competitors.

For now, the strategic imperative for businesses is clear: don’t ignore the rise of AI search. It may not replace traditional search overnight, but it is already reshaping how people find information and make decisions. Much as mobile browsing quietly grew until no business could afford to neglect it, AI-driven discovery is on a trajectory from novelty to normalcy. Companies that start addressing their “AI visibility” today – by understanding how they appear in AI-generated content and taking steps to improve it – will be better positioned for the future than those that wait until their web traffic mysteriously dries up.

In the early 2000s, savvy businesses learned to ask, “How do I rank on Google?” In the late 2020s, a new question is echoing in boardrooms and marketing meetings: “How do we rank on ChatGPT (or Gemini, or the next AI)?” The terminology might still be settling – whether we call it GEO, LLM search optimization, or something else – but the principle is the same. It’s about ensuring your brand has a voice in the new conversation. The web isn’t literally vanishing, but it is being distilled and repackaged in ways we haven’t seen before. To stay relevant, companies will need to make sure they haven’t vanished from that distilled view. In practice, that means mastering the art of being the answer (or at least part of it) when the question is asked to an algorithm. It’s a new kind of challenge, one that will likely define digital strategy in the coming decade. And for those who rise to it, the reward is not just surviving the transition, but thriving in a search landscape that’s starting to look less like an index of pages and more like an intelligent, if selective, conversation.