Study finds 29.8% of domains cited in Google AI Overviews never appear on page one
A Washington University study found 29.8% of domains cited in Google AI Overviews never appear on the first results page. Ranking well does not predict citation, and only 41.9% of overviews were fully grounded, per a dev.to summary.
The Engineer · Build desk
Key figures from the WashU study: 89% of 98,020 claims were clearly or broadly supported by their cited source; 41.9% of overviews were fully grounded, with every claim supported; the 11% error rate is an upper limit, as Reddit threads and YouTube videos were not fully captured.
| Figure | Value | Note | Claim |
|---|---|---|---|
| Claims clearly or broadly supported, of 98,020 | 89% | 11 | |
| Overviews fully grounded | 41.9% | 12 | |
| Error rate, an upper limit | 11% | Reddit, YouTube not fully captured | 13 |
What happened
- The researchers collected every AI Overview, its citations, the first-page results beside it and the content of each cited page across 55,393 searches between March 13 and April 21, 2026.
- Ahrefs, analyzing 863,000 keywords and 4 million AI Overview URLs, found 38% of cited pages in the organic top 10, down from 76% in July 2025.
- Overviews appeared on 64.7% of question-form queries and on 9.5% of non-question queries.
- Across 61,212 citations, cited domains scored higher on average for credibility than the first-page results shown for the same searches.
- Of 98,020 claims checked against their cited sources, 89% were supported, 7% were not addressed and 4% were directly contradicted.
Why it matters
- Who has to decide decision Teams optimizing for AI search need a citation log kept apart from the rank report: for each query, whether an overview appeared and which of their URLs it cited.
- What it limits constraint A citation log only has data on queries that trigger an overview, and those skew to questions and to topics like hobbies and science over politics and law.
- Who's exposed exposure A cited URL does not mean the overview represents the page correctly, so the log needs a spot-check of the claim sitting beside each citation.
Query fan-out is the mechanism that lets a page be cited without ranking. According to the dev.to summary of Google's documentation, the model splits the original query into related sub-queries, retrieves pages for each, and assembles the answer from the combined results [4]. A page that misses the primary query can still be cited if it answers one of the sub-queries [5]. In our view, a rank tracker that checks only the primary query never sees the sub-queries where that citation gets decided.
Read the 29.8% for what it counts. It is a share of cited domains missing from the first results page [1]. That puts about 70.2% of cited domains on page one [15]. The dev.to post then restates the finding as "Nearly a third of cited pages do not rank in the top 10" [2]. Domains on page one and pages in the top 10 are different tests.
Ahrefs counted pages, and its figure leaves 62% of cited pages outside the organic top 10 [19]. BrightEdge put the overlap at 17% [18]. The WashU and Ahrefs numbers cannot be averaged.
The sample behind the 29.8% is smaller than the search count suggests. At the study's overall trigger rate of 13.7% [6], 55,393 searches yield about 7,589 overviews [16]. The post does not break the 29.8% down by cause.
For the figure to transfer to your site, your queries would need a similar question-form and topic mix, and Google would have to retrieve the way it did during the study window. The second condition is the weaker one. The share of cited pages Ahrefs found in the top 10 fell to half its July 2025 level [20]. The post says the paper was presented at the ACM Internet Measurement Conference in October 2026 and calls it peer-reviewed [21]. That is at least five months after the data window closed [22].
The credibility result comes with a limit. The score is a composite, PC1, that accounts for domain authority, content quality and trust signals [10]. It does not isolate any one input, so a site owner cannot tell from it which one to change. Citations concentrate but leave a tail: the top 5 hostnames took 20% of them and the top 50 took 48.3% [14].
The researchers cautioned that the 11% error rate is an upper limit, because their pipeline could not fully capture dynamic content such as Reddit threads and YouTube videos [13].
What to watch
- Whether the full paper breaks the 29.8% down by cause, separating fan-out retrieval from any other path into the citation list.
- Whether Ahrefs' share of cited pages in the organic top 10 keeps falling after its drop from 76% in July 2025.
- Whether the researchers extend their claim-checking pipeline to Reddit and YouTube content, which would tighten the 11% upper limit.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence52
- Adoption
- Not scored
- Hype gap+25
- Incentives
- Not scored
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
29.8% of cited domains did not appear in the first-page results at all.
- [2]
The post states: "Nearly a third of cited pages do not rank in the top 10."
- [3]
The study analyzed 55,393 Google searches over 40 days (March 13 through April 21, 2026), collecting every AI Overview, its citations, the first-page search results displayed alongside it, and the content of every cited webpage.
- [4]
Per Google's documentation as relayed by the post, 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.
- [5]
A page that does not rank for the primary query can still be cited if it answers one of the fan-out sub-queries.
- [6]
13.7% of all searches in the study triggered an AI Overview.
- [7]
64.7% of question-form queries triggered an AI Overview, versus 9.5% for non-question queries.
- [8]
Topic affected activation: 46.1% of hobby and leisure searches and 39.9% of science queries triggered an overview, but only 7.5% of political searches and 9.6% of law and government queries.
- [9]
Across 61,212 citations, AIO-cited domains received higher average credibility scores than the traditional first-page results shown for the same searches.
- [10]
Credibility was measured by a composite metric (PC1) that accounts for domain authority, content quality, and trust signals.
- [11]
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, and 4% directly contradicted the cited source.
- [12]
Only 41.9% of overviews were fully grounded, meaning every claim was supported.
- [13]
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.
- [14]
An average overview cited about eight sources; the top 5 hostnames accounted for 20% of citations and the top 50 for 48.3%.
- [15]
About 70.2% of cited domains did appear in first-page results.
- [16]
At a 13.7% trigger rate, 55,393 searches yield about 7,589 AI Overviews, assuming the rate holds across the whole sample.
- [17]
Ahrefs analyzed 863,000 keywords and 4 million AI Overview URLs and found that only 38% of cited pages appeared in the organic top 10, down from 76% in July 2025.
- [18]
A separate BrightEdge analysis put the overlap at 17%.
- [19]
By Ahrefs' count, 62% of cited pages fell outside the organic top 10.
- [20]
Ahrefs' top-10 share of cited pages fell to half its July 2025 level.
- [21]
The post describes the study as peer-reviewed and presented at the October 2026 ACM Internet Measurement Conference.
- [22]
October 2026 is at least five months after the data window closed on April 21, 2026.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- dev.to29.8% of AI Overview sources don't appear in Google's first page at all
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Search Engine OptimizationFollow
- AI search citationsFollow
Entities
- GoogleFollow
- ACM Internet Measurement ConferenceFollow
- AI OverviewsFollow
- Washington University in St. LouisFollow
- BrightEdgeFollow
- AhrefsFollow