Short answer. An AI Overview is not a re-ordering of the organic top 10. Google decomposes a question into related sub-queries, retrieves candidate passages separately for each, and assembles the answer from whichever passages score best. In Surfer SEO's December 2025 study of 173,902 URLs, reported by Search Engine Land, about 68% of cited pages ranked in the top 10 for neither the main query nor any of its sub-queries. Ranking gets a page into the candidate pool; it does not decide who is cited from it.
Key takeaways
- Google's AI Overview splits one search into several related sub-queries — called query fan-out — and retrieves candidates separately for each one.
- Attribution happens at the passage level, not the page level, so a page can be cited for a paragraph on a topic its headline never mentions.
- Surfer SEO's December 2025 study of 173,902 URLs found about 68% of cited pages ranked in the top 10 for neither the main query nor any fan-out query.
- Pages ranking for the main query and at least one fan-out query were 161% more likely to be cited and accounted for 51% of all citations in that study.
- Published estimates of how often an AI Overview appears at all range from roughly 16% to roughly 50% of queries, depending on who measured and which query set they used.
How does Google's AI Overview decide what to cite?
It runs a retrieval pipeline against passages, not a quality judgment on whole pages.
Query fan-out is Google's practice of splitting one typed question into several related sub-queries — narrower specifications, rephrasings, translations, clarifications — before retrieving anything. Each sub-query then retrieves its own set of candidate documents, and the system chunks those candidates into passages: paragraphs, table rows, list items. It scores and ranks the passages rather than the pages they came from, then writes one answer from whichever passages served each sub-query best and attributes it to their sources. Because attribution happens per passage and per sub-query, a cited page can look unrelated to the results ranking below the AI Overview for the original question.
Why doesn't ranking on page one guarantee a citation?
Ranking in the top 10 for the main query is one entry point into the candidate pool, not the qualifier for being cited.
Surfer SEO extracted 33,000 fan-out queries with Gemini across a 10,000-keyword sample, then checked where the 173,902 cited URLs actually ranked. About 68% of cited pages ranked in the top 10 for neither the main query nor any fan-out query. The more useful half of the finding is about breadth: pages that ranked for the main query and at least one fan-out query were 161% more likely to be cited, accounted for 51% of all citations, and showed a Spearman correlation of 0.77 between the number of fan-out queries a page ranked for and its odds of citation. Breadth across a question's neighbourhood beats height on the question itself. Trackers also agree the pattern is widening, even where they disagree on the exact level — overlap between AI Overview citations and the organic top 10 has fallen sharply since 2024, with different trackers reporting different rates for the same period, which is a reason to treat any single overlap percentage as directional rather than exact.
What should a page publish to get cited?
A page earns citations by holding defensible, self-contained passages on the several questions adjacent to its main topic, not by ranking well for one keyword.
- Cover the question's neighbourhood on one page — definition, mechanism, exceptions, comparison, cost, limits — each under its own heading, each answerable without the others.
- Write headings as the sub-questions themselves. A fan-out query is a question a person would type; a heading phrased that way matches it directly, while a heading like "Our approach" matches nothing a retrieval system is looking for.
- Front-load the page. Industry compilations of AI Overview citation data, including Omnibound's, consistently find that a disproportionate share of citations come from early in the page rather than the closing sections, which argues for leading with the answer rather than building up to it.
- Keep passages self-contained. A paragraph opening with "it" or "this approach" cannot be lifted without its neighbour, so it is not lifted at all.
- Put structure where the facts are. Tables and lists are cleanly extractable units; prose with the numbers buried mid-sentence is not.
A page can gain citations while losing rank, and lose citations while holding rank, because the two are measured on different mechanisms. Reporting AI visibility against position tracking alone will describe a system that no longer exists — measure citations directly against a fixed question set instead.
Why do recently updated pages get cited more often?
Recency is the second-clearest citation signal after structure, and it is strongest on commercial and evaluation-stage questions.
According to AirOps' 2026 State of AI Search Report, cited by Omnibound, 83% of AI citations on commercial and evaluation-stage queries came from pages updated within the previous 12 months, and pages that go unrefreshed for a full quarter are markedly more likely to lose the citations they held. For most organisations that makes reviewing and genuinely updating an existing page a cheaper win than publishing a new one — a thorough page from 2023 that nobody has touched competes badly against a thinner page revised last month. Re-dating a page without changing its substance is a different act entirely, and it misleads more than the answer engine reading it.
How often does an AI Overview even appear?
Published trigger rates disagree enormously, from roughly one search in six to about half of all searches, because the rate depends heavily on the query set measured.
| Source | Trigger rate | Basis |
|---|---|---|
| Semrush, November 2025 | 15.69% of queries | — |
| Conductor, Q1 2026 | 25.11% of queries | 21.9 million queries |
| BrightEdge, February 2026 | ~48% | Commercial verticals |
| Google's own statement | ~50% | US queries |
The spread is a sampling artefact rather than a contradiction: informational, question-shaped queries trigger an AI Overview far more often than navigational ones, and Pew Research Center's tracking of real user searches found question-based queries ("who," "what," "why") produced a summary roughly 60% of the time versus far less for short, one- or two-word searches. The only trigger rate that supports a business case is the one measured across the actual questions a business's own buyers ask.
What is an AI Overview citation worth in traffic?
Less in direct clicks than a blue link used to be, and more in influence than a click-through number captures.
Pew Research Center's analysis of real Google search behaviour found users clicked a traditional organic result on just 8% of searches where an AI summary appeared, against 15% where none did — roughly half the click rate. Separate click-through-rate tracking compiled by Omnibound has shown similar queries-with-AI-Overview rates falling well below queries without one across 2024 and 2025, even as the exact figures moved from period to period. The honest framing is that a citation inside an AI Overview is worth much less in clicks than a ranked organic result, but worth something real in influence, because it is what the buyer actually reads before ever reaching a website. Build the business case on the second point rather than the first.
What can't be guaranteed about AI Overview citations?
Nobody can sell or submit their way into a citation, because there is no placement mechanism to buy.
- No one controls the citation. There is no submission process, no index request and no paid placement; anyone selling a guaranteed AI Overview citation is selling something they do not own.
- The fan-out set itself is unstable. Surfer SEO found only about 27% of extracted fan-out queries stayed consistent across repeated runs, which means optimising for one specific sub-query list is building on sand — optimising for topical breadth is not.
- A brand's own domain is one lever among several. Compilations of AI Overview citation data consistently find that most citations point at third-party sites rather than the brand's own domain, so off-site presence matters as much as the page itself.
- Coverage moves under everyone at once. Trigger rates have already been recalibrated more than once and will be again, so a measurement programme needs to survive that happening rather than assume today's rate is permanent.
The sibling surfaces behave differently again: Google AI Mode draws from a separate source list, ChatGPT retrieves from its own index, and Perplexity's citation behaviour is markedly more stable than Google's. None of the three can cite a page their crawler cannot fetch in the first place, which is the subject of AI crawler access and blocking.
How does Lifewood approach AI Overview visibility?
Lifewood treats AI Overview visibility as a measurement problem first, and a content problem second.
The instrument is a fixed question set per market, run repeatedly, with citations logged directly rather than inferred from rank tracking, because a page can gain citations while losing position and a blended report hides both movements. Content work then targets the fan-out neighbourhood around a topic rather than a single keyword — one page holding self-contained, evidence-carrying passages on definition, mechanism, comparison and limits, matching the pattern the difference between GEO, SEO and AEO sets out. The multilingual dimension is where delivery matters most, since a fan-out query asked in Vietnamese or Portuguese is rarely just the English question translated: Lifewood's 100+ languages and 40+ delivery centres across 30+ countries mean the question set and the content answering it are both produced in-market rather than translated afterward. See the AEO services and GEO services pages for how that work is delivered.