How to rank in AI Overviews: what 178 searches show
On 1 November 2026 this study pulled 178 everyday US searches and checked every page Google’s AI Overview cited against the ordinary results for the same search. Pages ranked higher were cited more often, but only one cited page in five ranked in the top 10 for that search, and four in five did not rank in the top 20 at all.
of the searches showed an AI Overview (156 of 178)
sources cited in each overview, on average
of cited pages ranked in the top 10 for the same search
of overviews cited the page that ranked first
Four in five cited pages were not in the top 20 for the search
Each overview cited about eight sources. For every one, the study looked for the same page in the organic results for that search, and separately for any page from the same site. Choose a group of searches below to see how the picture changes.
- ranked 1 to 3
- ranked 4 to 10
- ranked 11 to 20
- not in the top 20
What kind of source was cited
- YouTube
- Government sites
- Education sites
- Reddit and Quora
- Everything else
Across all 178 searches, 19.5% of cited pages ranked in the top 10 for the same search (95% interval 15.6 to 23.9%), 21.0% in the top 20, and 79.0% nowhere in the top 20. Half the overviews (52.3%) cited nothing from the top 20 at all.
The page ranked first was cited in a quarter of overviews
Share of overviews that cited the page ranked at each position, out of 153 overviews with sources. 1 November 2026, US, desktop.
The page ranked first was cited in 24.8% of overviews (interval 18.3 to 32.0%), and the pages at 9 and 10 in 10.5% and 8.5%. Higher positions were cited more often on the whole, though not in strict order: position 3 was cited less often than positions 4 and 6. No position was cited even a third of the time.
Matching by site mostly adds YouTube
Match by site instead of by page and the overlap rises: 47.7% of citations went to a site with some page in the top 10 (interval 43.9 to 51.5%). Much of that is YouTube. Every video counts as the same site, youtube.com, and videos make up 66% of the citations where the site ranked but the page did not. Leave YouTube out and the site figure is 37.7%, against 24.7% for the same page. The page ranked first had its site cited in 40.5% of overviews, or 36.0% once the 14 searches where a YouTube video ranked first are left out.
Google’s documentation suggests one possible reason for site-level matches. Its page on AI features and your website says AI Overviews may use a “query fan-out” technique, “issuing multiple related searches across subtopics and data sources”, and its guide to generative AI features says they are “rooted in our core Search ranking and quality systems”. That would fit pages from ranking sites being cited for a search they do not rank for. This study did not test fan-out, and Surfer, which simulated the fan-out searches, still found 67.82% of cited pages outside the top 10 for the search and its fan-outs.
A quarter of the citations went to YouTube
YouTube took 26.3% of all citations (interval 22.9 to 29.6%) and appeared in 72.5% of overviews. With google.com, mostly shopping links, Google’s own properties took 29.3%. In the how-to categories YouTube alone took more still: 48% of citations in home repair, 40% in software and 39% in food.
Government sites took 7.5% of citations, close to the 6% of AI Overview links to .gov sites that Pew Research Center found in 2025.
Reddit and Quora took only 1.9%, and appeared in 14.4% of overviews. That is far below the published lists of most-cited domains: Ahrefs’ monthly list for September 2026 puts Reddit second, with 18.5% of its top-50 total, and Pew found Wikipedia, YouTube and Reddit the most cited sources in AI summaries. Part of the gap is that forums were often missing from the top 10: Reddit or Quora ranked in the top 10 for 53 of the overviews, and 18 of those cited either, against 4 of the other 100. But even where a forum ranked, two overviews in three left it out.
The published studies disagree, and these figures sit low
The published figures disagree with each other as much as with these, partly because each counts something slightly different (the same page or the same site, the top 10 or the top 20, the plain blue links or every block on the page) and partly because their samples, dates and parsers differ. Almost all of them come from companies that sell AI-visibility tools. Each row below uses the same measure on both sides, as far as each study defines it.
| Study | Measure | Their figure | This study |
|---|---|---|---|
| Ahrefs, July 2025 | Cited pages that rank in the top 10, top three sources per overview | 76.1% | 21.2% |
| Ahrefs, March 2026 | Cited pages that rank in the top 10 blue links | 37.1% | 19.5% |
| Conductor, June to August 2026 | Cited pages that rank in the top 10 | 43% | 19.5% |
| BrightEdge, September 2025 and February 2026 | Cited sources that rank in the top 10 (page or site not stated) | 16.7% and about 17% | 19.5% |
| seoClarity, October 2025 | Citations from the top 20 | 56% | 21.0% |
| Conductor, June to August 2026 | Overviews citing the page ranked first | 24.9% | 24.8% |
| seoClarity, October 2025 | Overviews citing the page ranked first | 43% | 24.8% |
| Authoritas, August 2024 | Overviews citing the page ranked first | 66.5% | 24.8% |
| Kirsten and colleagues, 2025 (preprint) | Distinct sites each overview cites that have a page in the top 10, averaged per overview | about 47% | 40.6% |
These figures sit at the low end. They agree with BrightEdge on the top 10 and almost exactly with Conductor, the largest recent study available, on how often the number one page is cited. They are well below most of the others, old and new: the 2026 studies from Ahrefs and Conductor put the top-10 share at about twice this study’s. Ahrefs itself now reports about half its 2025 figure and says it has improved its parsing since then.
Two peer-reviewed papers point the same way. Grossman and colleagues (SIGIR 2026) found low overlap between the sources an AI Overview uses and the ordinary results for the same search. Sünkler and colleagues (WebSci 2026) found an AI answer on 87.7% of 900 US health searches, with 46.2% overlap between its sources and the ordinary results. The domain-level figure from Kirsten and colleagues is a preprint that has not yet been peer reviewed.
What about the clicks?
This study counts citations. For clicks, Pew followed the Google searches of 900 US adults in March 2025: people clicked an ordinary result on 8% of visits with an AI summary and 15% without one, and clicked a link inside the summary on 1%. Google says total organic clicks have been relatively stable, without publishing figures.
How it was done
- Searches
- 178 US English searches across ten everyday categories, taken from Ahrefs Keywords Explorer: the highest-volume matching terms for four seed words per category, each with at least 200 searches a month, with “near me” searches left out.
- Results pages
- One live Google results page per search from the DataForSEO API: United States, English, desktop, the organic results on the first two results pages (14 to 20 per search, 17.7 on average; “top 20” on this page means these), with the AI Overview loaded. All 178 were pulled between 06:27 and 06:33 UTC on 1 November 2026.
- Matching
- Same page means the same URL after ignoring http or https, www, query strings and a trailing slash, except Google Shopping links, which keep their query string because it names the product. Same site means the same registrable domain, so every YouTube video counts as youtube.com. Google returned most organic links wrapped in a redirect, and every one but one was resolved to its real address before matching.
- Intervals
- 95% intervals come from resampling the searches 2,000 times. Category figures rest on 4 to 22 overviews each, so treat them as a direction only.
- Limits
- One day, desktop only, US English. Searches phrased as questions showed an AI Overview almost every time (86 of 87), other searches 70 of 91. The 87.6% describes this sample only.
- Tools
- Ahrefs supplied the search list and DataForSEO the results pages. Both are paid data services. Ahrefs also publishes some of the studies compared above.
Download the data: the 178 searches (CSV) and the 1,268 citations (CSV), each with the cited page’s organic rank for the same search where it had one.
Sources
- Google Search Central, “AI features and your website”, updated 10 December 2025, and “Optimizing your website for generative AI features on Google Search”, updated 10 July 2026. Read 1 November 2026.
- Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, 22 July 2025.
- Google, “AI in Search is driving more queries and higher quality clicks”, 6 August 2025.
- Grossman and colleagues, SIGIR 2026, doi 10.1145/3805712.3809667. Sünkler, Lewandowski, Schultheiß and Koop, ACM WebSci 2026 Companion, doi 10.1145/3795513.3807429. Kirsten and colleagues, arXiv 2510.11560 (preprint).
- Ahrefs, AI Overview citation study, 21 July 2025, its update of 2 March 2026, and its monthly most-cited domains list, September 2026. Conductor Academy, AI Overviews against organic rankings, updated 25 September 2026. BrightEdge weekly AI search insights, 18 September 2025 and 12 February 2026. seoClarity research, updated October 2025. Authoritas, 27 August 2024. Surfer, AI Overview citation sources, updated 17 September 2026.
Want this checked for your own searches?
Book a call, tell me which searches matter to your business, and I can run the same check on them. A call takes about 30 minutes.
Book a call on CalendlyHow I work: GEO and AEO consulting for AI search, and SEO as a freelance SEO consultant.