Ahrefs' AI Overview Top-10 Citation Study, Digested

Ahrefs published the study on March 2, 2026, written by Louise Linehan. It asked one question: how many URLs cited in Google AI Overviews also show up in the top of the regular search results for the same query?

The answer matters because a lot of AI search advice assumes the two are the same thing. If AI Overviews mostly cite pages that already rank, classic SEO is the whole job. If they often cite pages that don't rank, organic ranking is only part of the picture.

This is a snapshot of Google AI Overviews as of March 2026. It says nothing direct about AI Mode, Gemini, ChatGPT or Perplexity. Google itself says AI Overviews and AI Mode may use different models and techniques, so a finding on one surface does not transfer automatically to the other.

Ahrefs compared the URLs cited in each AI Overview with the URLs on the same results page. A citation counted as an overlap when the exact cited URL also appeared in the results.

Two details shape how you should read the numbers:

• Result blocks versus organic results. The headline 37.9% counts the first 10 blocks on the page. A separate 37.1% figure counts organic results only.
• URL-level matching. The match is on the exact URL, not the domain. If Google cites one page on your site while a different page of yours ranks, the study scores that as no overlap.

That second point mechanically penalises large sites. A big publisher or retailer often has many pages on a topic. Google can cite a deep page while a category or hub page holds the ranking. The domain is present in both places, but the study records a miss.

Quoting the study directly: "37.9% of URLs cited in AI Overviews also appeared within the first 10 blocks." The organic-only figure was 37.1%.

The remainder splits roughly in two:

Read plainly: in this dataset, most AI Overview citations came from URLs outside the top 10. About a third came from pages that did not appear in the top 100 for that query at all.

That fits what Google documents about how its AI features retrieve. Google says it may use a query fan-out technique, issuing multiple related searches across subtopics. A page can be cited because it ranks for one of those hidden sub-queries, not for the query you typed. Google's AI features documentation also says eligibility needs only that a page is indexed and eligible to show with a snippet, with no additional technical requirements.

It does not show a decline

An earlier Ahrefs study put the top-10 share much higher, around 76%. It's tempting to read 76% falling to 38% as AI Overviews drifting away from rankings. The evidence doesn't support that. Ahrefs changed its parsing methodology between the two studies, so the two endpoints come from different instruments. The 38% figure stands. The trend does not.

It does not settle the number

BrightEdge (September 18, 2025) reported that only 16.7% of AI Overview citations came from the top 10, while also claiming 54.5% overlap. It published no sample size and did not define "overlap". Either one method is wrong or the two measure different things. They should not be cited as agreeing. This kind of conflict is common, and why GEO studies disagree comes down to definitions, samples and instruments more often than to the engines changing.

It does not show cause

Overlap is a correlation. A page can rank and get cited for the same underlying reasons, such as topical fit or brand prominence, without one causing the other. Large, well-known sites tend to do well on both, and the study has no way to control for that.

It cannot be checked against Google's own data

Google states that third-party tools don't have access to its internal ranking data. Its Search Console report for AI features shows impressions only, with no position. So any rank figure in a study like this comes from a third-party crawl, not from Google.

Rank is a moderate, decaying proxy for AI Overview citation. It's worth having, and it's not a guarantee. Here is what's in your control:

1. Keep doing solid SEO, but don't treat a top-10 position as the finish line. In this study, a ranking page was cited in a minority of cases. The guide on organic rankings and AI Overview citations covers how the two relate.
2. Check eligibility before anything else. A page must be indexed and snippet-eligible to be a supporting link. Look for stray noindex, nosnippet, max-snippet:0 or data-nosnippet that a template may have added.
3. Check the Search Console generative AI control. Inclusion is the default, but a property can be opted out, and child properties inherit from the parent.
4. Make sure the right page answers the question. Because matching is per URL, the page Google cites may not be the page you optimise. State the specific fact (a price, a definition, a number) on the page most likely to be fetched.
5. Measure AI Overviews per query, not as a rank. There is no position in an AI Overview to track, which is why free AI rank tracking can't give you a number Google would confirm. Check whether an overview appears and which URLs it cites.

Don't add markup expecting a lift. Ahrefs' separate schema and AI citations study found no major uplift, and AI Overview citations fell 4.6% on treated pages. Nobody controls which retrieved page Google finally cites.