Search Console queries show concepts the page never covers (RankBrain)

Google confirms RankBrain relates words in a search to concepts and ranks pages it judges related. Impressions for an uncovered concept suggest Google associates the page with it, but the page cannot satisfy it. This is a proxy: the check reads Search Console data, not RankBrain itself.

Decide whether the concept belongs on the page. If it does, add a clearly headed section that answers it. If it does not, create or link to the page that does, and tighten the page's title and introduction to its real concept.

1. In Search Console Performance, filter to one page and export at least 90 days of queries with impressions and average position.
2. Group the queries by the underlying concept they ask about, not by shared words.
3. For each group with meaningful impressions and an average position worse than 10, search the page text for any passage that addresses that concept.
4. Flag groups with no matching passage, and record whether the concept is in scope for the page.

Written for the RankBrain page (evidence: Confirmed). Google has not documented this system in a way a site can test directly, so this check measures a proxy for it, not the system itself.

Indirect. Estimate only: better positions for the affected query group where the concept is genuinely in scope; no effect where the association is noise.

Google's ranking systems guide and its 2022 post by Pandu Nayak describe RankBrain relating words to concepts; the link between uncovered query concepts and weak rankings is Laurelin's inference.