Somewhere today, a potential customer asked ChatGPT which product to buy in your category. They got a confident, specific answer with three or four brands in it. Either you were one of them or you weren’t, and nothing in your analytics will tell you which.
That is the uncomfortable reality of AI search in 2026. A growing share of buying journeys now starts inside a chat window or an AI-generated answer box: ChatGPT, Perplexity, Gemini, and Google’s AI Overviews sitting on top of the results page. These systems don’t return ten blue links. They return an answer, and the brands named inside that answer collect the demand.
The old scoreboard stopped working
For two decades, the discipline was simple: rank higher, get more clicks. The assumption carried over naturally to AI search, since everyone figured the AI would simply cite whoever ranked. For a while, that was roughly true. It isn’t anymore.
Ahrefs studied 863,000 search results and four million cited URLs in early 2026 and found that only about 38% of Google AI Overview citations now come from pages ranking in the organic top ten, down from roughly 76% in mid-2025. Even more striking, 31% of citations go to pages ranking beyond position 100, pages that effectively don’t exist in traditional search.
In other words, traditional rank and AI citation have decoupled. You can rank first and never be cited. You can sit on page nine and be quoted in answer after answer.
Why this happened: the fan-out
The mechanism isn’t a mystery, because Google has documented it. When you ask an AI-powered search a question, the system performs what Google calls query fan-out: it issues multiple related searches across subtopics and data sources, then assembles an answer from the passages it retrieves. The unit of competition is no longer the page ranking for a head term. It’s the individual passage that best answers one of those invisible sub-queries.
That explains how a page ranking beyond position 100 wins citations. It contained one paragraph that answered one fan-out sub-query better than anything else, and the system lifted it.
What “AI visibility” actually means
AI visibility is the discipline that grew up around this shift. It asks three questions that classic SEO never had to:
- Are you mentioned? When someone asks an AI assistant for recommendations in your category, does your brand appear in the answer at all?
- Are you cited? When AI systems do link to sources, are your pages among them?
- What is being said? AI systems will describe your pricing, your features, and your reputation whether or not the description is accurate. Brands routinely discover that ChatGPT confidently states pricing they retired two years ago.
None of these show up in a rank tracker, which is why most companies have no idea where they stand. The blind spot is the defining feature of the moment: the channel is growing, the answers are influencing buyers, and almost nobody is measuring their presence in it.
The currency changed too
One more finding worth knowing. In a separate Ahrefs study correlating eleven factors against AI visibility across 75,000 brands, branded web mentions showed the strongest correlation with appearing in AI answers, roughly three times stronger than backlinks. The systems that generate these answers learn about brands from text, wherever that text lives: reviews, comparisons, forum threads, podcast transcripts. A mention works on an AI engine even when it never carried a link.
That inverts a lot of conventional wisdom. The link-building budget that made sense in 2020 might do less for AI visibility than a steady drumbeat of honest mentions across the places where your category gets discussed.
What to actually do about it
The good news is that the playbook is becoming clear, and the early-mover advantage is real because so few companies act on it.
Measure before you optimize. Sample the questions your buyers actually ask, across the engines they actually use, and record whether you appear. Do it manually with a spreadsheet if you have to. Dedicated platforms exist for exactly this; LLMRanks for example tracks where a brand shows up across AI engines and what those answers say, then ties the findings to a concrete action plan. Whatever tooling you use, the point is the same: you cannot fix a presence you haven’t measured.
Publish something only you can publish. AI systems cite pages that add information they can’t get elsewhere: original numbers, real benchmarks, first-hand data. Google even ships an explicit reward for this now, labeling primary reporting as “Highly Cited” in its AI surfaces. Rehashed content gives an answer engine nothing to quote.
Write passages, not just pages. Because of fan-out retrieval, a page earns citations one self-contained passage at a time. Lead sections with the direct answer, keep each section able to stand alone, and cover the cluster of sub-questions around your topic rather than repeating one keyword.
Build mentions, not just links. Digital PR, podcast appearances, expert commentary, genuine community participation. The goal is your brand’s name appearing in trustworthy text, with or without a hyperlink attached.
The window
Every search shift has had a window where the discipline was cheap because nobody was practicing it. SEO in 2003. Social in 2009. AI visibility is in that window now: the answers are already influencing buyers, while most brands haven’t even checked what those answers say about them.
The brands that get named in AI answers this year are the ones the next year’s models keep treating as the default answer. Citations compound. Absence compounds too. Checking where you stand costs an afternoon, and it’s the rare marketing question where you might genuinely not know the answer until you ask.