How AI Overviews choose sources: what the public evidence says.

What AI Overviews are

AI Overviews are AI-generated summaries that appear above or below the standard search results for certain queries. They were rolled out across English-language results in the United States in 2024 and expanded through 2025 and 2026. They are different from featured snippets: featured snippets pull a single block from one page; AI Overviews synthesise multiple sources and present a written answer with named citations.

For an expert-led company, the practical question is whether their pages appear in those citations. The AI Citation Rate Index (AICRI) published earlier addresses that question generally.

What Google has said

Google Search Central's AI Overviews and your website page sets out a small number of named signals. Pages that follow Google's existing search guidance are eligible to be cited. There is no separate optimisation for AI Overviews.

Google's documentation also notes that AI Overviews cite pages for the same reasons that pages rank: relevance to the query, entity clarity, and demonstrated expertise.

Where Google has been less specific is the relative weight of citations versus rankings.

What the public evidence suggests

Across the audits we have run in 2024 and 2025, four patterns recur.

First, entity clarity is a prerequisite. AI Overviews tend to cite pages whose Organisation, Person, and Service entities are unambiguous and stable.

Second, sourceable evidence matters more than content volume. Pages that cite primary data, named methodology, or attributable case material are cited more often than pages that paraphrase common knowledge.

Third, URL persistence matters. AI Overviews appear hesitant to cite URLs that have recently changed or that sit behind redirect chains.

Fourth, cross-source corroboration helps. A page that is referenced by Wikidata, by an industry directory, and by a recognised professional association gets cited more often.

What is not known

Google does not publish a ranking factor list for AI Overviews. We do not know the exact weight of any of the signals above.

We also do not know whether AI Overviews handle commercial queries differently from informational queries.

Finally, we do not know whether Google's behaviour in the United States generalises to Europe, Latin America, or other markets.

What to do with this

If you run an expert-led company, the answer is not to optimise for AI Overviews. The answer is to follow the same entity clarity + sourceable evidence + URL persistence discipline you would follow for any other search surface, then re-measure quarterly.

If you are already running that discipline, AI Overviews are likely to surface your pages when they are relevant.