Does Ranking on Google Still Get You Cited in AI Answers? What the Data Shows
Many teams that already invest in SEO ask a version of the same question: "We rank well on Google. Does that carry over into AI answers, or do we need something on top?"
The short answer is that ranking is still the entry ticket, but it is no longer a reliable sign that AI answers will cite you. Large studies by the SEO software company Ahrefs show that most pages cited in AI answers do not rank in Google's top 10 for the question the user asked. Outside Google, the link is weaker still.
This article walks through that data, explains why rankings don't carry over, and sets out what an SEO team should measure and change as a result.
How often do AI Overviews cite pages that rank in the top 10?
Google's AI Overviews are the AI-written summaries shown above the regular search results. Each one links to a few supporting pages, which this article calls citations. The question is how many of those cited pages also rank on the first page of results for the same search.
Ahrefs has measured this twice. In July 2025, it found that about 76% of AI Overview citations came from pages ranking in the top 10 for that search. Among the top three citations in each Overview, the median cited page ranked third.
In March 2026, Ahrefs repeated the analysis on 863,000 searches and 4 million cited URLs. This time, only about 38% of cited URLs ranked in the top 10. About 31% ranked between positions 11 and 100. Another 31% did not appear in the top 100 at all.
Two caveats matter. First, Ahrefs improved how it detects citations between the two studies, so part of the drop from 76% to 38% may reflect better measurement rather than a change in Google's behavior. Second, AI Overviews vary from one run to the next, and both studies come from a vendor that sells an AI visibility product. The exact percentages should be treated as estimates.
The direction is clearer than the precise numbers. In the most recent data, most cited pages did not rank in the top 10 for the search the user typed.
What this means: If roughly six in ten AI Overview citations go to pages outside the top 10, a ranking report cannot tell you whether you are being cited.
What about ChatGPT, Gemini and Copilot?
For AI assistants, the gap is wider. In an August 2025 study, Ahrefs ran 15,000 long-tail prompts through ChatGPT, Gemini, Copilot and Perplexity. Long-tail prompts are longer, more specific questions. Ahrefs then compared the pages each assistant cited with Google's top 10 for the same prompt.
On average, only about 12% of cited URLs ranked in Google's top 10. For the citations inside ChatGPT's answers, the figure was 8%. Perplexity was the exception at about 29%.
Switching the benchmark to Bing doesn't solve the problem. The average overlap with Bing's top 10 was about 10%. Neither search engine's first page was a good guide to what the assistants cited.
This study is a snapshot from mid-2025 and covers long-tail prompts only. Still, it fits the AI Overview pattern, and it adds one more lesson: each engine behaves differently, so each one needs its own measurement. Our article on measuring AI visibility engine by engine looks at this in more detail.
Why don't rankings carry over?
If ranking doesn't predict citation, the next question is why. The leading explanation, offered by Ahrefs and consistent with Google's documentation, is query fan-out. Instead of searching only for what the user typed, an AI engine can run several related searches on subtopics, then build its answer from what those searches return.
Google's own documentation confirms that its AI features may use fan-out, "issuing multiple related searches across subtopics and data sources." When an engine searches for questions the user never typed, the page that ranks for the user's words may never be retrieved.
Ahrefs suggests that the pages appearing most often across these sub-searches are the ones that get cited. It also notes that the switch of AI Overviews to Google's Gemini 3 model in January 2026 may play a role. Both are interpretations, not measured causes, because Google does not show which sub-searches it ran. Fan-out may also not be the whole story. In its 2025 study, Ahrefs found that pages cited from beyond the top 10 ranked for fewer and shorter keywords than top-10 pages. It said this "doesn't neatly fit" the profile it expected from fan-out content.
Some engines do reveal their sub-searches, and they show how far the engine can drift from the user's words. In one conversation tracked by Genezio, an AI visibility platform, a buyer asked ChatGPT which AI SEO tool measures visibility in ChatGPT or Google AI Overviews, without naming any vendor. While answering, ChatGPT ran this sub-search: "AI SEO tools measure ChatGPT Google AI Overviews brand visibility enterprise Profound Scrunch Peec AI Otterly 2026."
ChatGPT added vendor names, the word "enterprise" and a year. A page that ranks for the buyer's original wording would not necessarily rank for that search. This is a single example, not a rate, but it shows the mechanism. Our guide to tracking query fan-out explains how to observe these sub-searches.
The type of source that gets cited also differs from a typical results page. Ahrefs reports that YouTube is the most-cited domain in AI Overviews in its data. YouTube links also made up 18.2% of the cited URLs that didn't rank in the top 100.
Genezio's tracking points the same way. For AI Overview answers to prompts such as "we already do seo. what do we need on top of it for chatgpt and ai overviews?", two of the three URLs most often listed as sources were YouTube videos. The most-cited URL was a roundup of tools. That is a small sample from one category, but it matches the larger dataset.
So does SEO still matter for AI visibility?
Yes. None of this data says that ranking hurts, or that SEO is obsolete.
Google states that to appear as a supporting link in AI Overviews or AI Mode, a page "must be indexed and eligible to be shown in Google Search with a snippet." It adds that there are no extra requirements and that "all existing SEO fundamentals continue to be worthwhile." A page that Google can't crawl or index can't be cited by its AI features at all. Our article on blocking AI training without disappearing from AI answers covers how crawl access affects this.
The numbers support SEO too. In Ahrefs' latest data, about 38% of AI Overview citations still came from the top 10. Pages in the top 10 are heavily over-represented among citations, so ranking still appears to help.
The real conclusion is narrower: SEO is necessary but not sufficient. Ranking makes a page eligible and appears to help its chances. It does not tell you whether you were cited, by which engines, or for which questions.
What should an SEO team change?
That gap is where generative engine optimization (GEO) comes in: the practice of measuring and improving how a brand appears in AI-generated answers. The evidence above suggests four practical changes. They are reasonable implications of the data rather than tested tactics.
Measure AI citations directly. Track whether each engine cites or recommends your brand for the questions your buyers ask, as a metric separate from rankings. Report it per engine, because overlap with rankings varies widely between them. Answers also change from run to run, so run each prompt more than once before trusting a number, and separate mentions from recommendations when you report results.
Compare your ranking pages with your cited pages. List the pages that rank for your priority searches and check which of them AI answers actually cite. Then look at the pages that get cited without ranking, whether yours or a competitor's. The differences show where your SEO work is and isn't carrying over.
Cover the sub-questions, not only the main query. Engines search related subtopics, comparisons and variants such as "enterprise" or the current year. Content that answers those questions directly gives the engine more to retrieve. Observed sub-searches are the best guide to which questions matter.
Look beyond your own website. Tool roundups, review articles and YouTube videos appear often among cited sources. Our guide to prioritising third-party sources covers where to focus.
Several AI visibility tools can support the first three steps. Tools such as Ahrefs' Brand Radar and Genezio track citations across several AI engines and show the fan-out queries behind a prompt. Genezio also tracks how competitors are cited and recommended across a market. Whatever tool you use, the principle is the same: collect citation data directly instead of inferring it from rankings.
The bottom line
Ranking on Google still opens the door to AI answers, but it no longer predicts who gets cited. Most pages cited by AI Overviews don't rank in Google's top 10 for the user's question. In Ahrefs' 2025 test of long-tail prompts, the vast majority of pages cited by AI assistants didn't either. Keep your SEO fundamentals, and add direct, engine-by-engine measurement of AI citations on top of them.