How Google's AI Overviews and AI Mode choose what to cite

Google says AI Overviews and AI Mode choose supporting links using the same core ranking and quality systems as regular Search. A page has to be indexed and eligible to appear with a snippet. After that, the model runs related searches and picks from what comes back. Google does not publish the selection rules beyond that.

What Google says a page needs

The most direct statement is in Google’s documentation on AI features and your website. To be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet. The page says there are no additional technical requirements, no special schema.org markup, and no new machine-readable or AI text files to create.

Google adds a caveat that is easy to skip. Meeting every requirement does not guarantee crawling, indexing, or serving. Eligibility is the floor, not a promise.

The same page notes that AI Overviews “often don’t trigger.” Google says they appear only when its systems decide the summary adds something beyond classic Search. So a missing Overview on a query you care about is normal.

Query fan-out changes which pages are considered

Both features can use what Google calls query fan-out: issuing several related searches across subtopics and data sources to build a response. In its guide to optimizing for generative AI features, Google gives an example. For the query “how to fix a lawn that’s full of weeds,” the fan-out queries might include “best herbicides for lawns,” “remove weeds without chemicals,” and “how to prevent weeds in lawn.”

The same guide describes the retrieval step as grounding: the system relies on core Search ranking to pull relevant, current pages from the Search index, then reads those pages to write a response with links that support it.

This is why ranking for your head term is not a reliable predictor of being cited. The pages that get read are the ones that rank well for the sub-questions, and you may never see those queries in a rank tracker.

Ahrefs measured something consistent with this. In an analysis of 863,000 keyword results pages and about 4 million AI Overview URLs, it found that 37.9% of cited URLs also appeared within the first 10 blocks of the results page. Another 31.2% appeared in positions 11 to 100, and 31.0% were beyond the top 100 blocks. Ahrefs compared that with roughly 76% in its July 2025 analysis and suggested Overviews now lean more on the results for fan-out queries than on the original results page. Treat that as Ahrefs’ reading of its own data. It comes from Ahrefs’ Brand Radar tool, and Ahrefs sells AI visibility tracking.

AI Overviews and AI Mode are separate lists

Google states that AI Overviews and AI Mode “may use different models and techniques, so the set of responses and links they show will vary.” Outside data agrees. Ahrefs compared 540,000 pairs of AI Mode and AI Overview responses to the same queries using September 2025 U.S. data. The two features cited the same URLs only 13.7% of the time, even though the answers were 86% similar in meaning.

If you track one and assume the other follows, you will misread your position. Doily lists Google AI Overviews and Google AI Mode as separate engines for this reason.

Controls you have over inclusion

Google says AI is built into Search, so the robots.txt rules for Googlebot are the control for how your site is crawled for Search. To limit what is shown from your pages, you use nosnippet, data-nosnippet, max-snippet, or noindex. If you add one and still see your content in an AI feature, Google says to confirm Googlebot can see the control (the URL Inspection tool shows the HTML it received) and to allow time for a recrawl, which it says can take anywhere from several days to several months.

Google-Extended is a different switch. Google describes it as a robots.txt token that governs whether content Google crawls can be used to train future Gemini models and to ground responses in Gemini Apps and in Grounding with Google Search on Vertex AI. Google states it does not affect inclusion in Google Search and is not a ranking signal. Blocking it does not remove you from AI Overviews or AI Mode.

What the cited sources tend to look like

Pew Research Center took the Google searches that 900 U.S. adults made in March 2025, reran them in April, and found 12,593 that produced an AI summary. A few of its findings are useful for expectations, with the limits that it recorded only up to 3 source URLs per summary and covers one month of 2025:

• 88% of summaries cited 3 or more sources, and only 1% cited a single source.
• Wikipedia, YouTube, and Reddit together made up 15% of the sources listed in AI summaries.
• 6% of the sources linked in AI Overviews were .gov sites, compared with 2% in standard results.
• The median summary was 67 words long.

Pew did not study why those sources were chosen, and the feature has changed since. Read it as a snapshot of what appeared, not a ranking factor list.

What to do with this

Google’s guide is blunt that creating unique, helpful content will likely matter more over the long run than any other suggestion in it. It separates “commodity” content, which restates common knowledge, from content with a first-hand point of view. Its example is that a first-hand review gives a perspective a summary of existing content cannot. It also warns that producing a separate page for every fan-out variation to manipulate responses falls under its scaled content abuse policy.

Practical steps that follow from the documentation:

1. Check that your important pages are indexed, crawlable, and not blocked from snippets by nosnippet or max-snippet settings you forgot about.
2. Write the page that answers the sub-questions around your topic, rather than one page per phrasing.
3. Keep Business Profile and Merchant Center details current, since Google says AI responses can draw on them for local and product results.
4. Use the Generative AI performance report in Search Console, which Google points to for seeing how your content shows up in these features.

For a working checklist, see our guide on writing pages AI engines cite and the explainer on how AI Overviews work.

What nobody outside Google can tell you

Google has not published how it weighs pages once they are retrieved, and it says no third-party tool has access to its internal ranking or AI systems. Studies like the ones above describe outcomes. They cannot tell you why your competitor was cited and you were not on a given day.

That gap is the reason to measure instead of guess. Sample the prompts and queries you care about on a schedule, record which pages are cited, and watch how the set moves. If you rank well and still do not appear, the pattern in ranks but invisible in AI is a good place to start diagnosing.