Ranking #1 on Google and being recommended by AI are now two different competitions, with different winners. Ahrefs analyzed roughly 15,000 real prompts across ChatGPT, Gemini and Copilot and found that only 8–12% of the links these engines cite also appear in Google's own top 10 results for the same query. That means a page can hold the top organic spot for years and still never come up when an AI engine answers the exact same question. Here's what the overlap data actually shows, why the gap exists, and what it means for how you measure — and chase — visibility today.
Only 8–12% of the links ChatGPT, Gemini and Copilot cite in their answers also appear in Google's top 10 organic results for the same prompt. Ahrefs reached that figure by running roughly 15,000 real prompts through the AI engines and comparing every cited URL against Google's own top 10 for the matching query — the same figure covered in the wider AI search statistics roundup on this site. Put differently: for every ten links an AI answer cites, roughly nine of them wouldn't have shown up if you'd just checked your Google rankings.
Proof: Ahrefs — AI vs Google citation overlap study (15,000 prompts).
Bottom line: a #1 Google ranking is no longer a proxy for AI visibility — it's evidence for a different, much narrower question.
Google and AI engines aren't measuring the same thing, so a low overlap is the expected outcome, not a glitch. Google's top 10 rewards keyword relevance and accumulated authority across a huge index of candidate pages. A generative engine instead has to pick a small handful of sources — often three to eight — that it can actually read, extract a clean claim from, and safely put in front of a user without contradiction. That favors pages that state an answer plainly and can be corroborated elsewhere, over pages that merely out-rank competitors on backlinks.
Do this: stop treating your Google rank as a stand-in for "AI already knows about us" — the two systems are shortlisting from different criteria entirely.
Some of the overlap gap isn't about content quality at all — it's that a chunk of high-ranking pages are never even reachable by an AI crawler, or aren't written in a form an engine can safely lift. Reflexa's own check of 40 well-known companies' robots.txt files found that 18% block at least one major AI crawler outright, which caps their AI visibility at zero regardless of Google rank. On top of that, the questions people actually type into ChatGPT or Claude tend to be longer, full sentences rather than the short keyword fragments used on Google — so a page written to match a keyword can miss the conversational question entirely, even while it ranks well for the keyword itself.
Proof: Reflexa — AI crawler access study (40 sites); arXiv — comparative study of AI vs. search query structure.
Rule of thumb: before asking why a top-ranking page isn't cited by AI, check the basics first — can the crawler reach it, and does it answer the question the way a person would actually ask it.
Because the overlap is this low, a Google rank-tracking tool tells you almost nothing about whether ChatGPT, Claude or Perplexity are naming you for the same buyer question. The two need separate measurement: SEO tools watch the ten blue links, but they don't run your actual buying questions against AI engines and record whether you were mentioned, recommended, or cited alongside your competitors. Without that separate check, a business can be confident about its search visibility and still be completely absent from the answers buyers are reading instead.
Shortcut: the free Reflexa check runs your real buying questions across ChatGPT, Claude and Perplexity and shows whether you're mentioned, recommended or cited — the measurement your Google rankings can't give you.
Citation-overlap studies vary by which AI engines are included, which prompts are sampled, and the query categories tested — Ahrefs' 8–12% covers ChatGPT, Gemini and Copilot on a broad prompt set, and other studies using different engines or narrower categories report different numbers. Treat the range as directionally right rather than a fixed constant. What's consistent across the data is the direction: the overlap is small, and it's not something you can infer from where you already rank.
Sources: Ahrefs — AI vs Google citation overlap (15,000 prompts) · Reflexa — AI crawler access study · arXiv — AI vs. search query structure
The free check runs the real buying questions on your domain — 3 minutes, evidence included.