AI search ranking factors: what we know

Google documents that AI Overviews and AI Mode draw on relevant web sources from its existing Search systems, and that pages must be crawlable, indexable and eligible to appear in normal Search. It does not publish a formula for which page gets cited. So the honest answer is that some factors are documented, many are only observed, and the exact selection rules remain unknown.

If you're trying to understand why a page appears in AI search, the hardest part isn't finding a list of ranking factors. It's separating what search engines actually document from what SEOs have observed.

That distinction matters. Google says its AI search features continue to rely on core Search systems and SEO fundamentals, but it does not publish a formula for deciding which page gets cited in an AI response.

What factors are actually supported?

The safest way to discuss AI search visibility is to separate evidence into four levels.

• Documented. Google says AI Overviews and AI Mode use relevant web sources from its Search systems. Pages need to be crawlable, indexable and eligible to appear in normal Search. Google also recommends helpful, reliable, people-first content, crawlable links, good page experience, textual content and appropriate structured data.
• Observed. SEOs can compare which pages are retrieved, cited or surfaced across repeated queries. These observations can reveal patterns, but they don't prove that a particular signal caused the result.
• Hypothesis. It is reasonable to think that factors such as topical relevance, source quality, clear entities or useful supporting evidence may affect retrieval or selection. But unless the search engine documents the mechanism, these remain interpretations.
• Unknown. Exact ranking weights, citation thresholds, model-specific selection rules and any supposed "AI citation score" aren't publicly documented. Treat claims about them cautiously.

The practical factors worth working on

Relevance

Relevance is the starting point. Search systems need to identify pages that address the user's question. Google describes Search results as being selected according to relevance and quality, using many factors rather than one universal signal.

That means a page targeting a broad topic isn't automatically useful for every related AI query. A page that directly answers a specific question, explains the conditions around the answer and covers the important subtopics may be easier to use as a source.

Useful content

Useful content is also directly supported by Google's guidance. Google asks whether content provides original information, useful analysis, substantial value and evidence of expertise rather than simply rewriting existing sources.

This is more actionable than trying to optimize for an imagined AI score.

Crawlability and indexability

These are prerequisites. Google states that pages need to be indexed and eligible to appear in Search to be eligible as supporting links in AI Overviews or AI Mode. It also recommends making important content available in text and ensuring crawling isn't blocked.

Freshness

Freshness matters when the query itself depends on changing information. A current pricing page, software documentation page or product comparison may need updating more often than an evergreen explanation. That doesn't mean changing publication dates or constantly rewriting pages. Google explicitly warns against changing dates simply to make content appear fresh.

Source quality and expertise

These are useful ways to think about why one page may be more useful than another. Google's guidance emphasizes original information, expertise, trust and substantial value. It does not say that an "authority score" determines AI citations.

Links and entities

Links and entities can help search systems discover and understand content. Google says links are used both to discover pages and as a relevance signal. Structured data can also help search engines understand what information on a page represents. But neither should be treated as a guaranteed AI citation mechanism.

Structured data

Structured data is best viewed as supporting machine understanding and eligible search features, not as an AI ranking shortcut. Google's AI search guidance specifically says there is no special schema.org markup required for AI Overviews or AI Mode.

For a deeper implementation discussion, see structured data for AI SEO.

Example: two pages answering the same question

Imagine the query: "Does FAQ schema help AI search?"

Page A gives a short definition, repeats common SEO advice and says FAQ schema "helps AI understand your content."

Page B answers the same question but distinguishes between structured data helping search engines understand page content and evidence for AI citation selection. It explains current Google support, notes that FAQ rich results have changed, and separates documented behavior from SEO interpretation.

Page B may be more useful to a search system because it gives a clearer, more specific answer supported by evidence. It also gives the reader fewer reasons to search again.

But that does not mean Page B will definitely be cited in an AI Overview or AI Mode response. Google does not publish a rule saying that this type of page receives a citation. That distinction is the important part.

If you need help improving this kind of page, read how to write citation-worthy content.

Try this on your site

Pick one existing article and spend 10 to 15 minutes auditing it.

• Choose one real query the page is intended to answer.
• Write down the exact answer the page gives in two or three sentences.
• Check the supporting evidence. Are claims backed by primary sources, examples or first-hand information?
• Look for ambiguity. Could a search system misunderstand what the page is actually about?
• Check crawlability. Make sure the important content is available to search engines and isn't hidden behind technical barriers.
• Compare the page with two competing results. Ask what useful information they provide that your page doesn't.
• Improve the missing part, rather than adding another generic section.

This is a better AI-search exercise than adding keywords or inventing a target citation score.

If you need a broader technical review, work through the AI SEO checklist.

What we still don't know

There is a strong temptation to turn observations into formulas: "AI citations reward X," "entities are worth Y," or "this schema increases citation probability."

There isn't enough public evidence to make those claims reliably.

Google has explained that AI Overviews and AI Mode can use multiple searches and sources to construct responses, and that the links shown can vary between AI features. That makes simple one-factor explanations especially unreliable. If you want the mechanism behind those multiple searches, see query fan-out examples.

A better working model is:

Then treat patterns you discover as observations, not universal ranking rules.

For a broader introduction to the subject, start with AI SEO for beginners, or follow the stage-by-stage path in how to learn AI SEO.

FAQ

Does Google have an AI citation ranking factor?

Google has not published a specific AI citation score or weighting system. It says AI features use Search systems and existing SEO fundamentals, but the exact selection process for individual supporting links isn't publicly documented.

Does structured data improve AI citations?

Structured data can help search engines understand page content and can support eligible search features. Google says there is no special schema required for AI Overviews or AI Mode, so structured data shouldn't be treated as a guaranteed citation tactic.

Should I optimize differently for Google AI Overviews?

Google's current guidance says the same fundamental SEO practices remain relevant. Focus on crawlability, useful and reliable content, clear information, good page experience and appropriate structured data rather than a separate AI ranking formula.

Conclusion

The most useful way to think about AI search ranking factors is not as a secret list of signals.

Some things are documented. Some are repeatedly observed. Some are reasonable SEO interpretations. Others remain unknown. That framework keeps AI SEO practical without turning assumptions into facts.