Recall
Recall is a retrieval metric measuring what fraction of all the relevant passages a system actually found, capturing whether the right content was retrieved at all.
Recall asks: did the system find the good stuff? It's the share of all truly relevant passages that a retrieval step managed to surface. High recall means few relevant sources were missed; low recall means good answers — possibly yours — never made it into consideration.
For AEO, recall is the first hurdle: if you're not retrieved, you can't be cited, no matter how authoritative you are. The levers that improve your odds of being recalled are being indexed (crawlable), using the real terms and concepts of the question, and writing clear, extractable passages that match intent. Recall is usually traded off against precision — casting a wider net finds more relevant items but also more noise, which reranking then cleans up.
Example. If ten pages genuinely answer a question and the engine retrieves eight of them, recall is 80%. The two it missed lose their shot at being cited — so being clearly matchable is what keeps you in the retrieved set.
Relevant pillar
Related terms
- PrecisionPrecision is a retrieval metric measuring what fraction of the passages a system returned were actually relevant, capturing how much of the result set is signal versus noise.
- RetrievalRetrieval is the step where an AI system searches an index to find the most relevant passages for a query before generating an answer, and it decides which content is even eligible to be cited.
- RerankingReranking is a second pass in retrieval where an initial set of candidate passages is reordered by a more precise relevance model, deciding which few actually make it into the AI's answer.