If your page ranks #1 on Google for a query today, there's a good chance it still won't show up as a citation in the AI Overview sitting above it. New data from Ahrefs — an analysis of 863,000 keyword SERPs and 4 million AI Overview citation URLs, published March 2, 2026 — found that just 37.9% of pages cited in AI Overviews also rank in the top 10 organic results for the same query. A year earlier, in July 2025, that overlap was 76%. For B2B SaaS, professional services, and B2B e-commerce brands that built content strategy around page-one rankings, the math underneath that strategy has changed.
The collapse tracks closely with one event. On January 27, 2026, Google made Gemini 3 the default model powering AI Overviews worldwide, expanding a rollout that had previously routed only the hardest queries to the new model. Gemini 3 brought stronger reasoning to AI Overviews — and, per a separate SE Ranking study of 100,000 keywords across 20 niches, it also replaced approximately 42.4% of the domains that had previously been cited, concentrated mostly in longer-tail queries where authority signals are thinner.
The mechanism behind the shift is one Google has publicly confirmed: query fan-out. Rather than answering your exact search phrase, AI Overviews silently expand it into a cluster of related sub-queries, then pull citations from whichever pages perform well across that expanded set — not necessarily the page that ranks #1 for the original keyword you were tracking.
Ahrefs ran the comparison two ways: once counting every SERP feature (ads, featured snippets, video packs, organic listings), and once isolating standard organic blue links only. The organic-only breakdown is the clearest read on what's actually happening to page-one rankings:
| Where the cited page actually ranks organically | Share of AI Overview citations |
|---|---|
| Top 10 (page one) | 37.1% |
| Positions 11–100 | 26.2% |
| Outside the top 100, or not ranking at all | 36.7% |
More than a third of AI Overview citations now come from pages that don't rank anywhere in Google's top 100 for the query being answered. A page-one ranking is no longer a prerequisite for AI citation — and, just as importantly, it's no longer sufficient on its own.
Consumer content has room to lose a few citations here and there. B2B SaaS and professional services pipelines usually depend on a much narrower set of high-intent commercial queries — "best [category] software for [use case]," "[competitor] alternative," "how to choose a [service] vendor." If AI Overviews are increasingly pulling citations from outside the traditional top 10 for exactly those queries, a page-one ranking stops functioning as reliable proof of AI visibility. It has to be tested and tracked directly, the same way you'd audit off-site citation signals rather than assume they exist because your own site looks strong.
The same SE Ranking study found Gemini 3 pulls roughly 32% more sources into each AI Overview response than the model it replaced. That tracks with a longer-running trend SE Ranking has tracked separately: AI Overviews now cite an average of 13.34 sources per response, up from just 6.82 in November 2024 — nearly double, with the maximum sources in a single response climbing from 25 to 95. More sources cited per answer means more competition for any single citation slot, but it also means Google is casting a meaningfully wider net than it used to, which cuts both ways for a brand trying to earn a spot.
Ahrefs' data also points to where the reshuffled citations are landing. YouTube is now the single most-cited domain in AI Overviews, up 34% over the past six months, and it accounts for 18.2% of all citations pulled from pages that don't rank in Google's top 100. For B2B categories where buyers watch product demos, comparison videos, and implementation walkthroughs, that's not a side note — it's a citation channel most B2B content teams still aren't producing for.
Seeing a competitor's page-one ranking slip and assuming your own AI Overview visibility is safe by comparison is a mistake. Ranking position and citation status are now two loosely correlated signals, not one. Each needs to be measured on its own.
Query fan-out means the unit of optimization has shifted from a single keyword to an entire cluster of related questions a buyer might ask along the way. SEO consultant Ethan Lazuk, who has written extensively on Google's fan-out documentation, put it this way:
"The more relevant you can make the passages in your documents (or other media) to fan-out queries, the likelier you'll be to earn a mention or a citation in the AI-generated answer. So for the question of how query fan-out impacts rankings, just know that we're no longer optimizing for individual keywords but rather entire user journeys, and those fan-out queries guide the way." — Ethan Lazuk, SEO Consultant
None of this makes page-one rankings worthless — they're still correlated with citation, just no longer synonymous with it. The brands adapting fastest are the ones treating AI citation as a metric to measure and manage in its own right, rather than an assumed byproduct of traditional SEO performance.
We'll show you exactly which of your pages are ranking but not getting cited — and what's actually earning citations in your category right now.
Book a Free AI Visibility CheckAccording to Ahrefs' analysis of 863,000 keyword SERPs and 4 million AI Overview citation URLs, just 37.9% of pages cited in AI Overviews also rank in the top 10 organic results for the same query as of March 2026. That's down from roughly 76% in July 2025, meaning a page-one Google ranking is no longer a reliable predictor of AI citation.
The drop tracks closely with Google making Gemini 3 the default model for AI Overviews on January 27, 2026. Gemini 3 leans more heavily on Google's confirmed "query fan-out" process, which splits a single search into multiple related sub-queries and pulls citations from whichever pages perform well across that expanded set, not just the original keyword.
Query fan-out is the technique Google has confirmed it uses to power AI Overviews and AI Mode: a user's search is expanded into several related sub-queries behind the scenes, and the AI-generated answer draws citations from whichever pages rank well across that broader cluster. It matters because content optimized for a single target keyword can miss most of the fan-out queries that actually determine citation.
It still helps, but it's no longer sufficient on its own. Ahrefs' data shows 36.7% of AI Overview citations now come from pages that don't rank in Google's top 100 at all for the query in question, and SE Ranking found Gemini 3 replaced roughly 42.4% of previously cited domains. Page-one ranking and AI citation should now be tracked as two separate signals.
The best-supported approach is to build content around the full fan-out cluster of related sub-queries rather than a single keyword, strengthen off-site presence on third-party sites Google's AI already trusts (forums, review sites, YouTube, industry publications), and structure content so individual passages can be extracted cleanly. Tracking actual citations directly — not just rank position — is essential to know if it's working.