How to rank in AI Overviews and show up in Google's answers
Contents · 12 sections
To show up in Google AI Overviews, a page has to be indexed and eligible to appear in Search with a snippet source. Ranking goes with citation, but it does not guarantee one. In TIN’s measurement of 69 US buyer searches, 30.7% of top-10 organic domains were also cited in the AI Overview, and 63.0% of the domains the AI Overviews cited were not in that search’s top 10 source. So rank for the question. Answer it better than the commodity pages. Publish where Google’s AI looks beyond page one, YouTube above all. Skip the schema and llms.txt tactics Google says it does not use.
These are associations in one snapshot, not causes, from one crawl window on buyer queries, not a random sample of all searches source.
How an AI Overview picks its sources, in Google’s own words
Google’s guide to its generative AI features, last updated 2026-07-10, covers both AI Overviews and AI Mode source. AI Mode is the sibling surface. Most of what follows applies to both, and where TIN’s data separates them, this guide says so.
The features run on ordinary Search. Google says its generative AI features “are rooted in our core Search ranking and quality systems” source. It names two techniques. Retrieval-augmented generation relies on those ranking systems “to retrieve relevant, up-to-date web pages from our Search index”. Query fan-out is “a set of concurrent, related queries generated by the model” that fetch more results for the user’s question source. TIN’s inference, not a measured finding: fan-out may explain part of the 63.0%, since a page can rank for one of the questions the model asks on the user’s behalf without ranking for the user’s own words.
Eligibility is a floor, and nothing guarantees a citation. A page “must be indexed and eligible to be shown in Google Search with a snippet”, and a site must also be included in Search generative AI features in Search Console source. Google adds that meeting every requirement still does not guarantee that a page is crawled, indexed or served source.
Google calls AEO and GEO plain SEO. “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO” source. Its 2025 blog post on the same topic gives the same advice in shorter form: unique content, a good page experience, crawl access, structured data that matches the visible page source.
Not every search gets an AI Overview, so not every keyword is worth this effort. On TIN’s own buyer searches, the crawler recorded an AI Overview on 75.8% of 91 usable US searches source. The feature itself has been pulled back before: after its May 2024 launch produced the glue-on-pizza answers, Google added triggering restrictions and limited user-generated content source.
What ranking on Google gets you
TIN read the Google results page and the AI Overview for 438 buyer searches in seven markets, in one crawl window on 2026-10-01 and 2026-10-02 source. The study was pre-registered, its changes are published as dated amendments, and its 1,760 rows are downloadable from the study page source. The US is the only market with enough searches for a headline.
“In the US, 30.7% of top-10 organic domains are also cited in the AI Overview (152 of 495, across 69 searches)” source. The share swings from search to search: “the median search shares 29%, the range runs from 0 to 86%, and 12 searches share no domain at all” source.
| Organic position (US) | Searches where that domain is also cited | Searches with a result at that position |
|---|---|---|
| 1 | 44.9% (31) | 69 |
| 2 | 46.4% (32) | 69 |
| 3 | 34.8% (24) | 69 |
| 4 | 26.1% (18) | 69 |
| 5 | 24.6% (17) | 69 |
| 6 | 19.4% (13) | 67 |
| 7 | 21.1% (12) | 57 |
Source for the table: TIN, Google top 10 vs AI Overviews.
Position 8 had too few searches to chart: “26 searches, 19.2%” source. And “top 10” is a label, not a count. “No US search returned a 9th or 10th organic result: SERP features take those slots, so “top 10” here means the organic results the page actually showed, up to 8” source.
Matching exact pages instead of domains gives a floor: “at least 25.7% of US top-10 organic pages (118 of 460) are the very page the AI Overview cites” source.
The overlap barely moves with the kind of query.
| US query type | Searches | Top-10 domains also cited |
|---|---|---|
| Choosing | 33 | 31.0% |
| Comparing | 14 | 30.6% |
| Pricing | 13 | 31.5% |
| Doing | 9 | 28.8% |
Source for the table: TIN, Google top 10 vs AI Overviews.
“Business queries share 30.2% (62 searches); consumer queries 35.4%, on only 7” source. The consumer figure rests on too few searches to mean much.
Most cited domains are not on page one
“63.0% of the domains US AI Overviews cite are not among that search’s top-10 organic domains (259 of 411)” source. That count “includes 35 opaque Google redirect links, which cannot be matched to any page and so count as outside” source.
The study’s own advice: “Track both, because one is not a stand-in for the other” source.
Why 76% and 63% are both right
Ahrefs published a study of 1.9 million citations from 1 million AI Overviews, and found that “76.10% of AI Overview-cited pages rank in the top 10” source. Set beside TIN’s 63.0% outside the top 10, that reads like a contradiction. It is two different measurements. Ahrefs has since re-run the study across 863K keyword SERPs and 4M AI Overview URLs, and found that “37.9% of URLs cited in AI Overviews also appeared within the first 10 blocks” source. The table compares the July 2025 study, whose method is the closer match to TIN’s.
| Ahrefs, July 2025 | TIN, October 2026 | |
|---|---|---|
| What is counted | Cited pages (URLs), the top 3 most visible citations in each AI Overview | Every distinct cited domain in each AI Overview |
| Compared against | Where the cited page ranks: top 10, positions 11 to 100, or below 100 | The organic results of the same search, which ran to 8 at most |
| Queries | 1 million AI Overviews; the query set is not described on the study page | 69 US buyer searches with an AI Overview |
| Headline | 76.10% of cited pages rank in the top 10 | 63.0% of cited domains are not in that search’s top 10 |
Sources for the table: Ahrefs; TIN, Google top 10 vs AI Overviews.
Ahrefs also reports that 9.50% of cited pages rank between positions 11 and 100, and 14.40% do not rank in the top 100 source. The study page reports where cited pages rank; it does not say in words whether that ranking is for the same search that produced the AI Overview.
TIN’s inference, not a measured finding: counting only the three most visible citations favours pages that also rank, while counting every cited domain, redirect links included, takes in the long tail of the answer. Which difference accounts for the gap has not been measured.
Where the two do meet is the top spot. In the same Ahrefs post, Si Quan Ong writes of ranking first that the chance of a citation “is a coin flip at best” source. In TIN’s study of Google top 10 vs AI Overviews, the US domain at position 1 was also cited in 44.9% of searches source.
What AI Overviews cite, by kind of source
A second TIN study asked seven AI engines the same 60 everyday questions in 10 countries and sorted every cited site into 10 categories source. It covers AI Overviews and AI Mode separately. Its sample is not the top 10 study’s: “60 questions are a sample of everyday searches, not of everything people ask AI. Business-to-business questions are not covered” source.
| Source category | AI Overviews (184 citations) | AI Mode (487 citations) |
|---|---|---|
| Company sites | 39.1% | 38.6% |
| Social and video | 14.1% | 11.1% |
| Associations | 9.8% | 8.8% |
| Government | 8.7% | 9.0% |
| Reference | 7.1% | 2.9% |
| Academic | 6.5% | 5.3% |
| News | 4.9% | 7.2% |
| Health providers | 4.3% | 2.1% |
| Comparison | 3.3% | 8.8% |
| Single-topic sites | 2.2% | 6.2% |
Source for the table: TIN, AI citation sources.
Company sites are the largest category in both source. The single site cited most is not a company site. In AI Overviews the most-cited sites were “youtube.com (17), en.wikipedia.org (8), britannica.com (3), heartfoundation.org.nz (3), isappscience.org (3)”; in AI Mode, “youtube.com (28), reddit.com (18), nerdwallet.com (7), canstar.com.au (5), en.wikipedia.org (5)” source.
Reddit is the split between the two surfaces: “Reddit is second in AI Mode (18 citations) but cited only 3 times in AI Overviews” source. AI Overviews cited 136 different sites, and their top 10 hold 26.1% of AI Overviews citations source. The two surfaces are close relatives, not twins: they “share 8 of their 25 most-cited sites” source.
Why this is not the same job as getting cited by ChatGPT
The same study puts ChatGPT next to Google’s AI answers, and the profiles barely overlap.
| Measure, citation sources study | AI Overviews | ChatGPT |
|---|---|---|
| Social and video, share of citations | 14.1% | 0.0% |
| Company sites, share of citations | 39.1% | 16.4% |
| Government, share of citations | 8.7% | 50.5% |
Source for the table: TIN, AI citation sources.
What to do, in order of the evidence
1. Be eligible, then rank for the question
Check the floor first: indexed, eligible for a snippet, and included in generative AI features in Search Console source. A page that fails any of the three is out before the answer is written.
Then rank. Positions 1 and 2 were the ones most often cited in TIN’s US data, in 44.9% and 46.4% of searches source, and the Ahrefs study found most of the citations it counted on pages ranking in the top 10 source.
2. Answer the question and its follow-ups, not the keyword
Google’s guide says content that people find “unique, compelling, and useful” will likely do more for a site’s presence in generative AI search than any of its other suggestions, and it contrasts that with commodity content “based on common knowledge” source. It also says you do not need to rewrite for AI: its systems “can understand synonyms and general meanings” source.
Fan-out is the reason to cover the follow-up questions on the page itself. But Google draws a line: creating separate content for every fan-out variation “primarily to manipulate rankings or generative AI responses” violates its scaled content abuse spam policy source. Google’s reason: “a high quantity of pages doesn’t make a website higher quality or more relevant to users” source. Ahrefs tested the fan-out theory on cited pages ranking beyond the top 10 and found they “actually appeared for fewer keywords, and returned for shorter queries”, which did not fit the profile it expected source.
3. Get talked about on the web, honestly
Ahrefs analysed 75,000 brands and correlated AI Overview visibility with brand signals for the roughly 74% that had any mentions source. Branded web mentions correlated at a Spearman 0.664 and backlinks at 0.218 source. The author warns that correlation is not causation source.
Google adds the boundary. Its features “can show what’s being said about products and services across the web, including in blogs, videos, and forum discussions”, but “seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem”, because its spam systems sit under the same features source.
4. Publish video where it answers the question
YouTube was the most-cited site on both surfaces source. The study’s own advice is direct: “For Google’s AI answers, publish where people watch and discuss” source. Google’s guide points the same way: its generative AI features “can bring in relevant images and video”, and following its image and video SEO guidance already counts as optimising for them source.
5. Keep pages crawlable and current
Google says “the way Google Search finds and processes your pages remains the core of how our AI systems access your data”, and that its AI models use “publicly accessible, crawlable content” source. Google states it can process content in JavaScript “as long as it isn’t blocked” source. Freshness comes from the retrieval step itself, which looks for “up-to-date web pages” source.
Preview controls count too. Google says nosnippet, data-nosnippet, max-snippet and noindex apply to its AI formats, and “more restrictive permissions will limit how your content is featured in our AI experiences” source.
6. Do not write to a length
Across 174,048 pages cited in 560,346 AI Overviews, content length sat at a Spearman 0.04 against citation position; 53.4% of citations went to pages under 1,000 words and 16.0% to pages over 2,000 source. Google says “There’s no ideal page length” source. Write until the question is answered, then stop.
What not to spend on
Google’s guide has a section called “Mythbusting generative AI search: what you don’t need to do” source. Set against the measurements, it reads like this.
| Tactic | What Google says | What the measurement says |
|---|---|---|
| Schema markup for AI | ”Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add” source | 1,885 URLs that added JSON-LD: AI Mode moved 2.4%, indistinguishable from zero; AI Overviews fell 4.6% relative to matched controls, a decline Ahrefs calls real but unexplained source |
| llms.txt and other AI files | Google Search “doesn’t use them”, and one “will neither harm nor help” visibility in Search source | Not measured for Google here; Google’s statement settles it for Google Search |
| Chunking content into small pieces | ”There’s no requirement to break your content into tiny pieces for AI” source | Not measured here |
| Rewriting content for AI | ”You don’t need to write in a specific way just for generative AI search” source | Not measured here |
| Bought or inauthentic mentions | ”isn’t as helpful as it might seem”; spam systems apply source | Real mentions correlate with visibility; nothing here measures fake ones |
| Tools claiming “internal” Google data | ”No third-party tool has access to our internal ranking or AI systems” source | Not applicable |
Schema still has a job: Google recommends it for rich results source. Keep it for that. If a guide sells schema as the AI Overview step, Google’s documentation disagrees with it.
How to check whether AI Overviews cite you
Start with Search Console. Google points to its Generative AI performance report “to measure how your content is performing in generative AI features on Google Search and Discover” source.
Then look at the answers themselves:
- List the searches you already rank for, positions 1 and 2 first.
- Read the live answer for each. Record whether you are cited, which page, and who is cited instead.
- Record the date, the market and whether you were signed in. TIN’s data is “a crawler’s view of the results page, not a real user’s personalised result” source, and a person sees a different page. A check with no date or market cannot be repeated.
- Check again later. TIN’s measurement is one crawl window, so one look at your own searches is an anecdote; repeat it.
- Keep AI Mode separate. In the source-mix study it cited 327 different sites from 487 citations; AI Overviews cited 136 from 184 source.
What this guide cannot tell you
- Correlations, not interventions. Both TIN studies say of their findings: “These are associations in one snapshot, not causes” source source. Nothing here shows that a change on your site causes a citation.
- Two different samples. The ranking overlap comes from buyer queries, “not a random sample of all searches” source. The source mix comes from 60 everyday questions; the study page describes its runs both as “one weekly run” and as “Two weekly runs”, and calls the result “a snapshot, not a trend” source. They are never one sample.
- The US headline rests on 69 searches source. In the source-mix study, AI Overviews returned no citations for France, Italy and Sweden source.
- One crawl window. The Google top 10 vs AI Overviews study is 2026-10-01 and 2026-10-02 only source. Google changes the feature; the 2024 retreat shows how fast.
- Vendor studies. The Ahrefs figures are printed with their method and their authors’ warnings.
Read next
- Google top 10 vs AI Overviews, TIN’s study: the full overlap data by market, position and query type, with the dataset.
- Which websites AI engines cite, TIN’s study: the source mix for all seven engines, including AI Mode.
- Generative engine optimization: the controlled evidence: the trigger rate, mentions and length studies in one evidence table.
- How ChatGPT ranks passages: the schema test in full, and how passages get retrieved.
- Google AI fails: what AI Overviews got wrong, what Google changed, and the Munich injunction.
- Google scaled back AI Overviews: the 2024 record of what the feature got wrong and what Google changed.
Get cited. If you want this measured on your own market’s searches, tell us what you sell.
Sources
- Google Search Central, “Optimizing your website for generative AI features on Google Search”, last updated 2026-07-10. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central Blog, “Top ways to ensure your content performs well in Google’s AI experiences on Search”, May 2025. https://developers.google.com/search/blog/2025/05/succeeding-in-ai-search
- TIN, “Google top 10 vs AI Overviews” study, edition 2026-10-02. https://theinternetninja.com/recon/google-top10-vs-ai-overviews-2026/
- TIN, “AI citation sources” study, edition 2026-10-02. https://theinternetninja.com/recon/ai-citation-sources-2026/
- TIN, “Generative engine optimization: what the controlled evidence shows”. https://theinternetninja.com/blog/generative-engine-optimization-what-the-controlled-evidence-shows-2026/
- TIN, “ChatGPT ranks passages, not pages”. https://theinternetninja.com/blog/ai-search-optimization-what-the-retrieval-record-shows-2026/
- TIN case file, “Google scaled back AI Overviews after it told users to put glue on pizza and eat rocks”. https://theinternetninja.com/stories/google-ai-overviews-glue-pizza-eat-rocks-scaled-back-2024/
- Ahrefs, “76% of AI Overview Citations Pull From the Top 10”, 21 July 2025. https://ahrefs.com/blog/search-rankings-ai-citations/
- Ahrefs, AI Overview brand visibility correlations, 75,000 brands. https://ahrefs.com/blog/ai-overview-brand-correlation/
- Ahrefs, short against long content in AI Overviews. https://ahrefs.com/blog/short-vs-long-content-in-ai-overviews/
- Ahrefs, schema and AI citations. https://ahrefs.com/blog/schema-ai-citations/
- Martinez, survey of studies on generative engine optimization, preprint. https://arxiv.org/abs/2607.14035
Questions
Do you need to rank on page one to be cited in an AI Overview?
Ranking goes with citation, but it is not required and it is not enough. In TIN's measurement of 69 US buyer searches, 30.7% of top-10 organic domains were also cited in the AI Overview, and 63.0% of the domains the AI Overviews cited were not in that search's top 10. Positions 1 and 2 were cited in 44.9% and 46.4% of searches. These are associations in one snapshot, not causes.
Does schema markup help you get into AI Overviews?
Google says structured data is not required for generative AI search and that there is no special schema.org markup to add, though it still helps with rich results. In an Ahrefs test of 1,885 URLs that added JSON-LD, AI Overview citations fell 4.6% relative to matched controls, a decline Ahrefs calls real but unexplained.
Do you need an llms.txt file for Google AI Overviews?
No. Google's guide says Google Search does not use llms.txt files or other special markup, and that creating one will neither harm nor help visibility in Google Search.
How can you see if your site is cited in AI Overviews?
Use the Generative AI performance report in Search Console, which Google describes as the way to see how your content performs in its generative AI features. Then read the live AI Overview for the searches you rank for, and record the date, the market and whether you were signed in.
Can you trick the AI Overview?
Google's guide says seeking inauthentic mentions is not as helpful as it seems, because its generative AI features depend on the same core ranking and spam systems as Search, and that creating pages for every fan-out variation primarily to manipulate responses violates its scaled content abuse policy.
How can I rank my website in AI search engines other than Google?
Treat each engine as its own channel. In TIN's study of seven engines on everyday questions, ChatGPT shared at most 1 of its 25 most-cited sites with any other engine, and it cited social and video platforms 0.0% of the time against 14.1% for AI Overviews.
Why does YouTube show up so often in AI Overviews?
In TIN's study of 60 everyday questions, youtube.com was the most-cited site in both AI Overviews and AI Mode, and social and video platforms took 14.1% of AI Overviews citations. Google's own guide says its generative AI features can bring in relevant images and video. Why YouTube leads is not measured; the study records associations in one snapshot.
Is SEO still worth it now that AI Overviews exist?
Yes. Google says its generative AI features are rooted in its core Search ranking and quality systems, and in TIN's US measurement the domains at positions 1 and 2 were also cited in 44.9% and 46.4% of searches, an association in one snapshot rather than proof of cause.
- Written by
- TIN research desk. No single byline: the brand is the work.
- Data used
- 2 studies, 1 case file, each linked where it is used: AI Overview citations: how often is a Google top-10 site also cited? 30.7% in the US, AI citation sources: which sites do AI engines cite? 2,857 citations from 7 engines in 10 countries.
- Rules
- Every figure carries its source link, and nothing is rounded or invented. How TIN works →
- Updated
- 2026-10-16