What Does the Shift in AI Overview Citations Mean for SEO?

What changed in AI Overview citations?

The tight link between ranking in the top 10 organic results and getting cited in Google AI Overviews has broken down over the past eight months. Ahrefs research, reported by Search Engine Journal on March 2, 2026, found that the share of AI Overview citations coming from pages already ranking in the top 10 fell from 76% in a July 2025 baseline to 38% in the March 2026 update, based on an analysis of 863,000 keywords and more than 4 million cited URLs.

The remaining citations are not clustered just below page one either. In the same dataset, 31.2% of cited pages ranked in positions 11 to 100, and another 31.0% ranked beyond position 100 organically or did not rank in the tracked organic results at all. Roughly one in five citations outside the top 100 pointed to YouTube, which alone accounted for 5.6% of all AI Overview citations in the dataset.

A separate, more recent study from WhatsMyGeoScore, published July 16, 2026 and covering 8,500 AI generated answers across ChatGPT Search, Perplexity, Claude, and Google AI Overviews, put current top-10 overlap at 17% to 54% depending on query type, with a correlation coefficient of just 0.34 between organic rank improvement and citation share increase. A weak correlation like that means moving up in rank still helps, but it explains a shrinking fraction of whether a page gets cited.

A third data point, from BrightEdge's 16-month analysis running through September 2025, found roughly 17% top-10 overlap using its own methodology. The exact numbers differ by research group and measurement window, which is normal for a fast-moving metric. What does not differ is the direction: three independent studies, over three different periods, all show page-one rank explaining less of AI citation behavior than it used to.

Who does this shift affect?

It affects anyone whose AI visibility plan was really just their SEO plan with a new label on it. If the working assumption was that a number one ranking guarantees a citation, or that years of accumulated authority keep a brand cited by default, that assumption now has a shrinking amount of data behind it.

B2B SaaS and tech growth teams competing for 'best [category] tool' prompts feel this directly. A page that has held the top organic spot for that keyword for years is now sharing citation space in ChatGPT and AI Overviews with pages ranked 40th, 60th, or 90th, simply because those pages answer a sub-question the AI system decomposed the prompt into.

Local, multi-location, and service businesses face a version of the same gap, and per the WhatsMyGeoScore breakdown below, local queries show the widest split between rank and citation of any query type.

There is an upside for challengers. Weaker correlation between rank and citation means newer domains without a decade of backlink history can win citation share faster than they could ever have won rank. LeadHaste, Omnicite's own operating case, went from 0 impressions to 1 million impressions, 200+ AI citations, and a Domain Rating increase from 1 to 24, in 4 months, without first spending years climbing to page-one rankings across its category.

Why did the top-10 correlation break down?

The main driver is query fan-out. Modern AI systems break a single prompt into a set of sub-questions and retrieve sources for each fragment separately, so a page that ranks first for the literal keyword may still miss several of the sub-questions the model is actually answering, while a lower-ranked page that covers one of those fragments in more depth gets cited instead.

The second driver is that there is no longer one index to satisfy. ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews each run their own retrieval process. A top-10 Google ranking says very little about standing inside Perplexity's or ChatGPT's separate retrieval layer, which is part of why the WhatsMyGeoScore study tracked all four engines rather than Google alone.

The third driver is format. The Ahrefs data reported by Search Engine Journal found YouTube responsible for 18.2% of citations that came from beyond position 100, and 5.6% of all AI Overview citations overall, meaning video is now a measurable citation asset competing directly with classic web pages, not a side channel.

How should you respond right now?

Stop treating organic rank as a proxy for AI visibility and start measuring citation share directly, meaning the percentage of relevant AI answers in a category that actually cite you. That is a different number from rank, and per the data above, the gap between the two is widening, not closing.

A few concrete adjustments follow from the same data:

1. Build topic clusters that cover the full fan-out of sub-questions around a subject, instead of a single page optimized for one keyword.
2. Diversify formats inside each piece: a comparison table, a sourced stat, an FAQ block, and where it fits, video, since format variety widens the surface area an AI system can quote from.
3. Ship precise structured data (FAQPage, HowTo, Article schema) so engines can parse and quote content cleanly rather than guessing at structure.
4. Track citation share, citation count, and answer presence across ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews, not Google position alone.
5. Keep publishing cadence current. Content that stops updating loses citation share to newer coverage faster than it loses organic rank.

What do the citation numbers look like by query type?

Top-10 overlap with AI Overview citations is not uniform. WhatsMyGeoScore's July 2026 breakdown, covering 2,100 informational, 2,800 commercial, 1,900 transactional, and 1,700 local queries, found a different overlap rate for each query type:

1. Informational queries: 38% to 47% of citations traced to top-10 pages.
2. Commercial queries: 45% to 54%.
3. Transactional queries: 52% to 61%, the highest overlap of any category.
4. Local queries: 29% to 41%, the lowest overlap of any category.

What does this mean for the next SEO or GEO review?

It means the review cadence itself needs to change. A shift from 76% to 38% inside eight months is too fast for an annual SEO audit to catch in time. Content and citation performance now need to be checked on a schedule closer to monthly than yearly, with citation share tracked as its own line item alongside rank.

None of this argues for abandoning SEO fundamentals. Ranking still feeds the retrieval process that produces citations in the first place. It argues for adding a citation layer on top of the rank layer, so that when the next study shows the correlation moving again, the response is a number you already track, not a surprise.

Sources

AI Overview citations from top-10 organic pages fell from 76% (July 2025) to 38% (March 2026), based on 863,000 keywords and 4 million+ cited URLs Search Engine Journal (reporting Ahrefs research), 2026-03-02

Top-10 overlap with AI Overview citations ranges 17% to 54% depending on query type, with a 0.34 correlation between organic rank improvement and citation share increase, across 8,500 AI generated answers WhatsMyGeoScore, 2026-07-16

A 16-month analysis through September 2025 found about 17% of AI Overview citations overlapped with top-10 organic pages BrightEdge, 2025-09-18