For twenty years, the answer to "how do I get found" was the same: rank. Everything else followed from that.
New data says that link has broken. In March 2026, Ahrefs analysed 863,000 SERPs and 4 million AI Overview citations and found that only 37.9% of cited pages ranked in Google's top 10. Nine months earlier, the same study put that figure at roughly 76% (Ahrefs).
It halved. Which means the majority of pages Google now cites in its own AI answers are pages Google does not rank on page one.
Ahrefs' analysis of 4 million AI Overview citations across 863,000 SERPs, published 2 March 2026 using Ahrefs Brand Radar. It is more than double the size of their July 2025 analysis, which covered 1.9 million citations. Both figures are quoted below so the change is visible.
What did the study actually find?
Of the pages cited in AI Overviews that appear anywhere in Google's results, the distribution is close to even thirds.
37.9% ranked in the top 10. 31.2% ranked between positions 11 and 100. 31.0% did not rank in the top 100 at all (Ahrefs, March 2026).
Filtering to blue links only, excluding ads and SERP features, the picture skews further from the top: 37.1% in the top 10, 26.2% at positions 11 to 100, and 36.7% beyond the top 100 (Ahrefs).
| Google position | July 2025 | March 2026 |
|---|---|---|
| Top 10 | ~76% | 37.9% |
| Positions 11 to 100 | Not broken out | 31.2% |
| Not in top 100 | Not broken out | 31.0% |
| Citations analysed | 1.9 million | 4 million |
Why would Google cite a page it does not rank?
Because an AI Overview is not answering the query you typed. It is answering a set of related questions it generated from your query, and it retrieves sources for each of those.
Ahrefs attributes the shift to exactly this: Google is selecting far fewer pages straight from the original SERP than it was in July 2025, drawing instead on expanded query results (Ahrefs). Your page does not need to be the best answer to the searched phrase. It needs to be the best answer to one of the sub-questions Google decomposed that phrase into.
This also explains why cited pages sometimes appear nowhere in normal results but do appear in SERP features (Ahrefs). The retrieval pass and the ranking pass are looking for different things.
“38% of pages cited in AI Overviews also rank in the top 10, down from approximately 76% in the previous analysis.
Is this true beyond Google?
It is more extreme elsewhere. Across AI platforms generally, Ahrefs found 80% of sources cited by AI search platforms do not appear in Google at all, and only 12% match Google's top results (Ahrefs).
So Google's own AI product is the most ranking-aligned of the major engines, and even it now pulls the majority of its citations from outside page one. ChatGPT, Perplexity and Claude are further away still.
If your AI visibility reporting is derived from your rankings, it is measuring a proxy that has largely stopped working.
What kind of page gets cited instead?
Three characteristics show up repeatedly in the data.
Freshness. AI platforms cite content that is 25.7% fresher than the content ranking organically (Ahrefs). Recency is weighted more heavily in retrieval than it is in ranking, which is why a well-maintained page can be cited while an older, stronger page ranks above it.
Format. YouTube makes up 18.2% of non-ranking citations, 5.6% of all AI Overview citations, and grew 34% in six months (Ahrefs, March 2026). A video result that has no traditional ranking to lose is now a meaningful share of what Google cites.
Brand presence. The strongest correlation Ahrefs found with AI visibility was branded web mentions, at 0.664 (Ahrefs). Not backlinks, not keyword targeting. How often your brand is credibly talked about across the web.
That last one has an encouraging counterpart. Yext studied 6.8 million AI citations and found 86% came from brand-managed sources: 44% first-party websites and 42% business listings (Yext). The pages being cited are overwhelmingly pages the brand controls.
Does this mean SEO no longer matters?
No, and the data does not support that reading.
Google still sends around 345 times more traffic to websites than ChatGPT, Gemini and Perplexity combined (Ahrefs). Ranking remains where nearly all the traffic is. A top-10 position also still triples your citation likelihood relative to your share of the index, since 37.9% of citations come from ten results out of potentially millions.
The correct conclusion is narrower and more useful: ranking is now necessary for traffic and insufficient for citation. They have become two jobs. Ranking work gets you clicks. Citation work gets you into the answer that replaces the click. Our breakdown of AEO versus SEO covers where the two overlap and where they genuinely diverge.
If you only do SEO, you will rank and be quoted less often than your ranking suggests. If you only do AEO, you will be cited while a competitor takes the clicks. Most businesses need both, weighted toward wherever their buyers actually are.
What does being cited get you?
More than the click you lost, if you are the one cited.
Seer Interactive found brands cited inside an AI Overview earn 35% more organic clicks and 91% more paid clicks (Seer Interactive). Against that, on queries where an AI Overview appears, organic CTR falls 61% overall, from 1.76% to 0.61%, and 83% of those searches end without a click (Seer Interactive).
Ahrefs measures the position-one hit at 58% fewer clicks, comparing December 2023 with December 2025 across 300,000 keywords (Ahrefs).
Read together, these say the same thing from two directions. AI Overviews are taking clicks from the page and redistributing a portion of them to whoever is inside the answer. Being cited is not a consolation prize for lost traffic. It is where the traffic went.
What should you actually change?
Four things follow directly from the data above.
Stop treating rank as your AI visibility metric. With 62% of citations coming from outside the top 10, your ranking report tells you increasingly little about whether you are being quoted. Measure citations directly. Our guide to checking your business in AI search covers how.
Answer the sub-questions, not just the headline query. Google is decomposing queries and retrieving per fragment. A page with clear, self-contained answers under question-shaped headings gives it more surfaces to retrieve. This is the mechanism behind most of what works in getting content cited by AI assistants.
Keep pages current. A 25.7% freshness gap is large enough to be a strategy on its own. Updating an existing page is often better value than publishing a new one.
Invest in being mentioned, not just linked. At a 0.664 correlation, branded mentions are the strongest lever in the dataset, and they are earned through presence rather than technical work.
What this looks like in practice
Consider a query like "best accounting software for small business".
Under the old model, Google ranked ten pages and users clicked one. Winning meant being one of those ten. Under the current model, Google decomposes that query into sub-questions it thinks the user actually wants answered: which options exist, what they cost, which suits sole traders, what the migration effort is, which integrates with a given bank feed. It retrieves sources for each fragment and assembles an answer.
Your comprehensive 5,000-word guide might rank second and be cited for none of it, because every sub-answer inside it is buried in context and depends on the surrounding paragraphs to make sense. Meanwhile a page ranking at 47 that answers the sole-trader question cleanly in four sentences under a matching heading gets cited, because that fragment stands alone.
This is why the ranking correlation weakened. The unit of competition changed from the page to the passage, and most content is structured for the old unit.
Take any section of your best page. Copy it out on its own, with no surrounding context. Does it still make sense and fully answer a question someone would ask? If not, it is unlikely to be cited, no matter how well the page ranks.
What this does not mean
Three overcorrections are worth naming, because each one is a plausible misreading of the same data.
It does not mean rankings are worthless. Ten results out of an index of billions supply 37.9% of citations. Being in the top 10 remains enormously over-representative, it has simply stopped being the whole story.
It does not mean you should publish thin pages. The pages being cited from beyond position 100 are not thin, they are specific. Depth on a narrow question beats shallowness at any length.
It does not mean traditional traffic stopped mattering. Google still sends around 345 times more traffic to websites than all AI platforms combined (Ahrefs). If you abandon SEO to chase citations, you will win the citation and lose the business.
Are you being cited, or just ranking?
Book a free AI Visibility Audit. We check where you actually appear across ChatGPT, Google AI Overviews, Gemini, Perplexity and Claude, and where your competitors appear instead.
Book Your AI Visibility AuditFrequently Asked Questions
How many AI Overview citations come from pages that rank in the top 10?
37.9%, according to Ahrefs' March 2026 analysis of 4 million citations across 863,000 SERPs. That leaves 62.1% coming from pages ranking at position 11 or worse, including 31.0% from pages not in the top 100 at all.
Why has this figure dropped so sharply?
The prior study, from July 2025, put it near 76%. Ahrefs attributes the change to Google selecting fewer pages directly from the original SERP and drawing more on expanded query results, meaning it retrieves sources for the sub-questions it generates rather than the query as typed.
Does this mean I should stop doing SEO?
No. Google still sends roughly 345 times more traffic than all AI platforms combined (Ahrefs), and top-10 pages remain heavily over-represented among citations. Ranking is still where the traffic is. It has simply stopped being sufficient for citation on its own.
What is the single strongest signal for getting cited?
In Ahrefs' data, branded web mentions, with a 0.664 correlation to AI visibility. Content freshness matters too, with cited content running 25.7% fresher than organic results.
Can a page with no Google ranking be cited by AI?
Yes, and it happens often. 31.0% of AI Overview citations came from pages outside Google's top 100, and across AI platforms generally, 80% of cited sources do not appear in Google at all (Ahrefs).
Ashur Homa
Built and scaled a digital brand to $100M+ in sales with zero ad spend. Has helped businesses generate millions through AI go-to-market strategy. Leads growth at Omni Eclipse.
Connect on LinkedIn
