The assumption that AI Overviews cite the same pages that rank organically is wrong, and the gap is wider than most SEOs think.
A peer-reviewed study by researchers at Washington University in St. Louis, presented at the October 2026 ACM Internet Measurement Conference, analyzed 55,393 Google searches over 40 days (March 13 through April 21, 2026). They collected every AI Overview, its citations, the first-page search results displayed alongside it, and the content of every cited webpage.
The finding that should change how you think about AI search optimization: 29.8% of cited domains did not appear in the first-page results at all.
What this means
Google uses a source-selection process for AI Overviews that is distinct from its conventional search-ranking system. The pages that get cited are not simply the pages that rank. They are pages retrieved through a different mechanism that draws from a wider pool.
Google's own documentation explains part of this. AI Overviews use query fan-out: the model splits the original query into multiple related sub-queries, retrieves pages for each, and assembles a response from the combined results. A page that does not rank for the primary query can still be cited if it answers one of the fan-out sub-queries.
The Ahrefs data confirms this independently. Their analysis of 863,000 keywords and 4 million AI Overview URLs found that only 38% of cited pages appeared in the organic top 10, down from 76% in July 2025. A separate BrightEdge analysis put the overlap at 17%.
The trajectory is clear: the relationship between organic ranking and AI citation is weakening, not strengthening.
The activation numbers
The WashU study also measured how often AI Overviews appear at all:
- 13.7% of all searches triggered an AI Overview.
- 64.7% of question-form queries triggered one, versus 9.5% for non-question queries.
- For non-question searches, activation rose from 9.9% for one-word queries to 38.7% for queries of six words or more.
- Topic mattered: 46.1% of hobby and leisure searches triggered an overview, 39.9% of science queries, but only 7.5% of political searches and 9.6% of law and government queries.
The pattern is clear: people who ask questions in natural language are the most likely to encounter AI-generated answers. Keyword-stuffing that worked for traditional SEO is less likely to trigger an AI Overview at all.
The credibility finding
The study scored domain credibility across 61,212 citations. AIO-cited domains received higher average credibility scores than the traditional first-page results shown for the same searches.
This is the part that should interest anyone producing content. The AI is not picking the most-marked-up pages. It is picking the most credible ones. Credibility, in this study, was measured by a composite metric (PC1) that accounts for domain authority, content quality, and trust signals. It is not a single factor you can optimize with a file or a tag.
An average overview cited about eight sources. The citation distribution was long-tailed: the top 5 hostnames accounted for 20% of citations, the top 50 for 48.3%, and the remaining 43% spread across 7,379 hostnames. There is room in the tail for sites that are not YouTube or Reddit, but the content has to be credible and specific enough to surface through fan-out retrieval.
The accuracy problem
The study decomposed AI Overview responses into 98,020 atomic claims and verified each against its cited source:
- 89% were clearly or broadly supported.
- 7% were not addressed in the cited text.
- 4% directly contradicted the cited source.
- Only 41.9% of overviews were fully grounded (every claim supported).
The researchers cautioned that the 11% error rate is an upper limit, since their pipeline could not fully capture dynamic content like Reddit threads and YouTube videos. But the finding is still significant: in a meaningful share of cases, the citation does not support the claim.
What this changes about SEO strategy
If you are optimizing for AI search, the data says three things:
Ranking for the primary keyword is not sufficient. Nearly a third of cited pages do not rank in the top 10. Covering a topic across related angles and sub-questions matters more than holding a single position.
Question-form queries are the primary surface. 64.7% of question queries trigger an AI Overview versus 9.5% of non-question queries. If your content answers specific questions clearly, it is more likely to be retrieved through fan-out.
Credibility is the signal, not markup. AIO-cited domains score higher on credibility, not on Schema density. Original research, first-hand experience, and expert analysis are what move this metric. Google's own guide says the same thing: "non-commodity content" with a "unique point of view."
The era of optimizing for a single keyword and a single ranking position is closing. The data does not say ranking does not matter. It says ranking is no longer the only path to being cited. The second path, through fan-out retrieval, rewards breadth and specificity over position.
Top comments (1)
We need to produce a short comment, one or two sentences, starting with lowercase, about this video specifically. Should mention reaction or question about the content. Maybe ask about methodology, or remark about surprising number. Must be casual, no formal. No hashtags, no quotes, no URLs. Use ASCII quotes only if needed. Avoid em-dash. Let's craft: "i didn't expect almost 30% of AI overviews to be missing from page 1, how do you pull those hidden sources?" That's good. Ensure lowercase start