Searches for how to rank in AI Overviews return long lists of tactics. The usual list runs to schema, an FAQ block, a short answer under every heading, and more links. Few of those lists say which items Google has actually confirmed.
Google has a page for this. It is short, and it says more about what a site cannot set than about what it can. This guide sorts the advice by who made the rule, then shows how much the outcome moves even when everything is done correctly.
Key takeaways
- Google states two requirements for appearing as a supporting link: the page is indexed, and it can show a snippet in Google Search. It says there are "no additional requirements."
- Google also states that no special schema, AI text file, or new markup is needed. Advice that treats markup as the lever goes beyond what Google says.
- Preview controls (nosnippet, data-nosnippet, max-snippet, noindex) are the one direct control Google documents. They limit what is shown. Google does not describe them as a way to raise a page's chances.
- Meeting every requirement does not guarantee a place in an AI Overview. Google says indexing and serving are not guaranteed, and that AI Overviews often do not appear at all.
- In a Zumi study of one buyer question, 77% of the domains AI Overviews cited appeared in a single answer and never returned.
What does Google say a site needs?
The source is AI features and your website on Google Search Central. It covers AI Overviews and AI Mode together, and its position is plain: the best practices for SEO remain relevant, and there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
The two stated conditions are technical. A page must be indexed and eligible to be shown in Google Search with a snippet. Google adds that meeting them does not mean it will crawl, index, or serve the content.
The same page lists ordinary fundamentals as still worthwhile: crawling allowed in robots.txt, internal links, a good page experience, important content in text form, and structured data that matches the visible text. None of it is specific to AI Overviews.
What is a control, a limit, or neither?
Sorting by who set the rule makes the advice easier to weigh.
| Type | What it is | Source |
|---|---|---|
| Requirement | The page is indexed and snippet-eligible | Google states it |
| Stated unnecessary | Special schema, AI text files, new machine-readable markup | Google states it |
| Direct control | nosnippet, data-nosnippet, max-snippet, noindex limit what Search shows from a page | Google states it |
| Not guaranteed | Crawling, indexing, and serving, even when every requirement is met | Google states it |
| Not controllable | Whether an AI Overview appears for a query, and which pages it links | Google: shown only when "additive to classic Search," so they "often don't trigger" |
| Observed, not confirmed | Answer-first writing, original data, brand mentions across the web, freshness | Third-party studies and practice |
The last row is where most published advice sits. Some of it is sensible. None of it comes from Google as a ranking rule, and it is worth holding to a different standard than the rows above it.
Does schema markup help a page appear?
Many guides recommend FAQ markup and other structured data as a way into AI Overviews. Google's page says the opposite: there is "no special schema.org structured data that you need to add."
That does not make markup pointless. Structured data still supports the Search features it was built for, and Google asks that it match the visible text on the page. The narrower point is that markup added only to appear in AI Overviews rests on a promise Google has not made.
Do pages need to rank in the top 10?
Third-party studies do not agree on how often cited pages rank in the organic top 10. BrightEdge, looking at AI Overviews a year after launch, reported that about 17% of sources cited in AI Overviews also rank in the organic top 10, and about 53% rank in the top 100.
Google offers one reason the picture is loose. It says AI Overviews and AI Mode may use a "query fan-out" technique, running multiple related searches to build a response. A page can be absent from the first results page for the question asked and still be cited for a related sub-question.
The practical reading is modest. Organic strength helps a page be considered, and it does not decide the outcome.
Do brand mentions across the web matter?
Ahrefs analyzed 75,000 brands and found that branded web mentions correlate with appearing in AI Overviews at 0.664, against 0.218 for backlinks. That is a strong association, and Ahrefs itself notes that correlation is not causation.
It supports spending effort on being discussed by credible third parties. It does not show that any given mention will produce a citation.
How much do the cited pages change with nothing changed?
A lot. Zumi asked one buyer question about choosing an AI search intelligence platform every day for 19 days. The domains AI Overviews cited each day overlapped with the day-one sources by 7%, and 77% of the domains it cited appeared in only one answer.
A separate Zumi run measured movement in brand mentions. Across 7,280 consecutive-day answer pairs on ChatGPT, Perplexity, and Google AI Overviews, AI Overviews changed least, at 0.54 on a scale where 0 means the same brands and 1 means none shared, against 0.64 for Perplexity.
Two limits apply. The citation study rests on 25 AI Overviews answers to one question, and the movement figure comes from one market, India beauty. Both describe how much the outcome moves, not why a page was cited.
What follows from that?
Three conclusions fit the evidence.
- The stated requirements come first. A page that is blocked, not indexed, or set to show no snippet cannot be a supporting link. This is the only part with a clear pass or fail.
- Preview controls are a deliberate choice. They are the one documented way to limit what Search shows from a page. They apply to Search generally, so the effect on classic results should be checked before they are set.
- Everything else is an input with no promised return. Clear writing, original data, and credible mentions are reasonable investments. A single citation, or its absence, is a weak signal about any of them.
How should the result be tracked?
Because the outcome moves on its own, a one-time check says little. Tracking a brand in AI Overviews and AI Mode covers the setup: a fixed question set, a separate baseline for each surface, repeated runs, and mentions counted apart from links.
Google's own reporting adds context. Search Console now includes generative AI performance reports, which show which pages appeared in AI features but not what the answer said about the brand.
Zumi, an AI Search Intelligence Platform, tracks Google AI Overviews on every plan and reports each engine on its own. Book a demo to see how a brand appears in the answers that matter to it.
Sources
- Google Search Central, "AI features and your website", last updated December 2025.
- Zumi, Gemini, AI Overviews, and AI Mode cite differently, July to August 2026.
- Zumi, how to evaluate AI visibility tools, India beauty question run, 23 July to 2 August 2026.
- BrightEdge, "AI Overviews at the One-Year Mark", February 2026.
- Ahrefs, AI Overview brand correlation study, 75,000 brands.