Ranking in the top 10 no longer wins an AI Overview citation: only about 38% of the pages Google cites in AI Overviews now rank in the top 10 for that same query, down from roughly 76% in Ahrefs' July 2025 study (the 2026 update covers 863,000 SERPs and about 4 million AI Overview URLs). Position and citation have come apart because they are scored on different units — a rank measures the whole page, while a citation is awarded to a single passage inside it.
The split is easy to miss until you watch it on real stores. Through Contexta we keep seeing WooCommerce pages parked at position 3 or 4 with steady impressions and almost no AI citations, sitting right next to page-two URLs that get quoted word for word. Because Contexta reads a site's own Search Console data, the gap it surfaces isn't a ranking problem — it's a structure problem, and that difference is the whole story behind the 76-to-38 drop.
Why does ranking in the top 10 no longer guarantee an AI Overview citation?
Because AI Overviews cite passages, not pages, and a page's rank barely predicts whether any single passage inside it answers a question on its own. In Ahrefs' 2026 data, only 37.9% of AI Overview citations came from pages in the top 10 for the query; the rest split almost evenly between positions 11–100 and pages ranking beyond position 100.
Read that distribution again: a page ranking nowhere near the top is now about as likely to be cited as one sitting on page one. BrightEdge, using a different methodology, puts the top-10 overlap even lower — around 17% as of February 2026 — so the exact figure is disputed, but every dataset points the same direction. The fact worth planning around is that winning the blue-link race for a query now settles a little over a third of who gets quoted above it. The rest is decided by something rank doesn't measure — how your content breaks into liftable pieces, which is a matter of passages, not word count or position.
Is this just query fan-out, or something deeper?
Fan-out is the visible mechanism; the deeper change is the unit AI Overviews actually retrieve. Google fans a query into several sub-queries, pulls candidate passages for each one, and scores each passage on whether it answers that sub-query by itself — so the citation is decided per passage, per sub-query, not per page, per head term.
That reframes the whole shift. A page ranking first for the head term can be invisible to a sub-query it never resolves in a self-contained chunk, while a page ranking 40th that nails that sub-query in one clean passage gets lifted instead. The overlap didn't fall because Google started preferring weaker pages; it fell because Google stopped citing pages and started citing passages, and rank was always a page-level number. Google's move to Gemini 3 as the global default for AI Overviews in late January 2026 is widely cited as an accelerant, but the underlying change in unit predates it. Featured snippets still behave the old way — about 99.58% of them come from the top 10 — which is precisely why the two surfaces now diverge: the snippet is picked from ranked pages, the AI citation from scored passages.
Can a page stuck on page two actually get cited by AI?
Yes — 31.2% of AI Overview citations in Ahrefs' 2026 study came from pages ranking 11–100, so page two and beyond is dead for clicks yet very much alive for citations. A URL Google buries at position 14 for the head term can still own the single best passage for one of the fan-out sub-queries, and that passage is what surfaces in the Overview.
We watched this on a store page ranking 12th for its category term: it earned no featured snippets and a thin trickle of organic clicks, yet an AI Overview quoted its "how to choose" paragraph because that paragraph answered a sub-query completely in two sentences. Nothing about its rank changed — what made it citable was that one section standing on its own. This is the counterintuitive upside of the collapse: the same shift that unseats your best-ranked pages hands your buried ones a way in they never had under classic SEO.
If position isn't the lever, what makes a passage citable?
A citable passage is a self-contained answer to one real question: a question-shaped heading, the answer stated in the first sentence or two, one concrete fact, and no reliance on the paragraph above it to make sense. That structure — not the page's rank — is what a retriever scores when it decides what to quote, and it holds whether the page sits at position 2 or position 92. The reusable shape is the answer-first block: heading, then answer, then detail.
Before you rewrite anything, size the prize: the click gap between a cited and an uncited brand is 0.18 percentage points, not a rescue, which means this work only pays on pages that already carry impressions. The harder half is knowing which of your ranking pages this is already costing you. The tell is a page with search impressions but a thin trickle of clicks — it's being reached and passed over. Contexta's Problem Map imports your Search Console data and ranks pages by lost clicks per month, pulling the CTR losers and page-two URLs to the top, so the pages that rank without being cited stop hiding inside an averaged report. From there the work is structural: recut each stranded section into a passage that resolves one query on its own.
So should you stop chasing top-10 rankings?
No — ranking still matters, it's just no longer the finish line. The top 10 remains the single largest citation bucket at about 38%, a strong position feeds the candidate pool a passage gets pulled from, and a page that can't be crawled or ranked won't be cited from anywhere. The "SEO is dead" takes overshoot: rank buys eligibility, it just stopped buying the citation outright.
What genuinely changed is where the decisive work sits. Under classic SEO you optimized a page to climb a ranked list; under AI Overviews you optimize each passage to answer a specific question a reader — or Google's fan-out — is actually asking. That's the same sentence-level move that separates GEO from SEO: keep earning the rank, but treat every section as a standalone answer, because as of mid-2026 the citation goes to the best passage, not the best position.
FAQ
Does ranking still matter for AI Overview citations in 2026?
Yes — ranking still matters, it just stopped guaranteeing a citation. The top 10 is still the single largest citation bucket at about 38% of AI Overview citations (Ahrefs, 2026), and a strong rank feeds the candidate pool a passage gets pulled from. But roughly 62% of citations now go to pages ranking outside the top 10, so a good position buys eligibility, not the citation itself.
Why did top-10 overlap with AI Overview citations drop from 76% to 38%?
Because AI Overviews shifted from citing whole pages to citing individual passages scored against fan-out sub-queries. Google splits a query into sub-queries, retrieves candidate passages for each, and picks the passage that answers that sub-query on its own — a decision rank never measured. Ahrefs recorded the top-10 share falling from ~76% in July 2025 to ~38% in 2026, with Google's move to Gemini 3 as the default AI Overviews model in late January 2026 widely cited as an accelerant.
Can a page that ranks on page two get cited by AI Overviews?
Yes — 31.2% of AI Overview citations in Ahrefs' 2026 study came from pages ranking 11–100, so page two is dead for clicks but alive for citations. A page Google buries at position 14 for the head term can still own the best passage for one of the fan-out sub-queries, and that passage is what surfaces in the Overview. What makes it citable is a self-contained section, not its rank.
How is a passage-based AI citation different from a featured snippet?
A featured snippet is chosen from pages already ranking in the top 10 — about 99.58% of them, per Ahrefs — while an AI Overview citation is chosen from passages scored against sub-queries, regardless of the page's rank. That's why the two surfaces now diverge: winning the snippet still requires a top ranking, but winning the AI citation requires a self-contained passage that answers one question completely, which a lower-ranked page can also supply.
