ObsurfableFor years, the practical rule was simple: rank in Google's top 10 and you had a good shot at appearing in AI Overviews. That rule broke in early 2026.
In March 2026, Ahrefs published an updated study analyzing 863,000 keyword SERPs and 4 million AI Overview URLs. The headline finding: only 37.9% of pages cited in AI Overviews also appeared in the top 10 organic results for the same query. In Ahrefs' prior study from July 2025, that figure was roughly 76%.
Google switched AI Overviews to Gemini 3 as the default model on January 27, 2026. The timing aligns with the collapse. Something fundamental changed in how Google's AI surfaces select sources — and the data points to query fan-out pulling from a much wider pool than the original SERP.
What the numbers actually show
Ahrefs broke cited URLs into three buckets:
| Organic position | Share of cited URLs |
|---|---|
| Rank in top 10 | 37.1% |
| Rank 11–100 | 26.2% |
| Outside top 100 | 36.7% |
Read that last row again: more than a third of AI Overview citations come from pages that don't rank in Google's top 100 for the query that triggered the overview.
Among citations from pages outside the top 100, 18.2% were YouTube URLs. YouTube alone accounted for 5.6% of all AI Overview citations in the dataset — often appearing in answers for how-to, health, and product queries where the video didn't rank organically.
Separate analyses tell a consistent story. BrightEdge put the top-10 overlap as low as 17% depending on methodology. SE Ranking found Gemini 3's arrival replaced roughly 42% of previously cited domains in a 100,000-keyword sample.
The global leaderboard of most-cited domains barely moved — YouTube, Reddit, Facebook, Indeed, Quora still dominate. What changed is which publisher and brand sites get cited within each query. B2B publishers and mid-market brands got reshuffled.
Why fan-out explains the reset
Gemini 3 brought the same agentic retrieval pattern already visible in AI Mode: query fan-out. Instead of lifting sources from the primary SERP, the model decomposes a question into multiple sub-queries, retrieves across Google's index (and beyond), and synthesizes one answer.
That means a page can earn an AI Overview citation because it is the best answer to a sub-question — even if it never ranked for the headline query. It also means YouTube, Reddit, and niche authority sites enter the citation pool more often, because fan-out retrieves entity-rich, experience-driven content that traditional ranking undervalues.
This is not "SEO is dead." Roughly 37% of citations still come from the top 10. Organic authority remains a strong signal. But rank is no longer a reliable proxy for AI citation — which is why teams that only track SERP position are missing half the picture.
What this means for your strategy
1. Monitor AI citations per prompt, not just rankings.
Define the questions your buyers actually ask — "best CRM for startups," "how to integrate X API," "is Y worth it" — and track whether your brand and URLs appear in the cited sources. A page-one ranking with zero AI citation is now a common outcome.
2. Treat YouTube as an AEO channel, not a brand-awareness afterthought.
YouTube is the most-cited domain in AI Overviews according to Ahrefs Brand Radar data, with citation share growing 34% over six months. Videos with clear titles, transcripts, and entity-consistent descriptions give fan-out another surface to pull from.
3. Build topic clusters for sub-queries, not single-keyword pages.
Fan-out rewards depth. Cover the sub-questions a complex query decomposes into — comparison, pricing, implementation, alternatives — so at least one page in your cluster matches each retrieval path.
4. Don't abandon SEO — but don't stop there.
Crawlability, E-E-A-T, and strong organic pages still feed retrieval. They are necessary, not sufficient. Pair SEO with extractable answer formatting, structured data, and presence on third-party sources models trust.
5. Re-measure after every major model change.
Gemini 3.5 Flash became the default for AI Mode at Google I/O 2026. Each model upgrade can shift citation patterns independently of your blue-link rankings. Treat model launches as trigger events for a full AI visibility audit.
How Obsurfable fits
Obsurfable runs your defined prompts against ChatGPT with web search and tracks brand mentions, competitor positioning, and cited sources over time. When a shift like the Gemini 3 citation reset happens, you see whether your AI presence moved — not whether your SERP position did.
The Visibility Director can flag gaps where competitors appear in AI answers and you don't, then draft content designed to close them.
For the broader Gemini model context, see What Gemini 3.5 Changes About AI Search Visibility. For fundamentals, see What Is Answer Engine Optimization?.
FAQ
Did Ahrefs compare identical methodologies across both studies?
Ahrefs notes its parsing improved since July 2025, so the two datasets are not perfectly comparable. Some of the drop may reflect better citation detection. But the directional shift is confirmed by BrightEdge, SE Ranking, and independent observers — the overlap genuinely collapsed.
Should I stop investing in SEO?
No. Top-10 pages still account for 37% of citations. SEO produces crawlable, trustworthy content that answer engines retrieve. The change is that SEO alone no longer predicts AI visibility.
Why is YouTube cited so often without ranking?
Fan-out retrieves the best match per sub-query. A well-structured how-to video can answer a sub-question better than any organic page, even when that video doesn't rank for the primary keyword.
The bottom line
The Gemini 3 citation reset is one of the clearest data points in AEO: top-10 organic rank went from a strong predictor to a coin flip. Fan-out retrieval, YouTube's rise, and sources outside the top 100 are the new normal. Measure what AI systems actually cite — because your rankings won't tell you.