Selection Rate · Share of times an AI picks a source
Selection Rate is the share of times an AI engine keeps a given source among all the candidate sources it retrieved for a query. It is the metric behind Selection Rate Optimization: the equivalent of click-through rate, applied to the model’s decision.
Standalone definition · 42 words
How is a source’s Selection Rate calculated?
Divide the number of times the source is used by the number of times it appeared among the candidates. Dejan defines it as the share of grounding results actually incorporated in the answer. You need to see the candidates, not only the final citations.
Selection Rate vs citation rate vs Citation Share: what is the difference?
Selection Rate is about the step where the model picks its sources. Citation rate, as Peec AI measures it, compares a URL’s citations to its retrievals. Citation Share compares your citations to competitors’. They overlap but measure different things.
What raises a source’s Selection Rate?
In Dejan’s tests, a brand that is strong in training data gets a good rate even with average content, while a fuzzy brand struggles even when its page is retrieved. Passage form then matters: direct relevance to the sub-query and extractable sentences.
Frequently asked questions about Selection Rate
Does a good Selection Rate guarantee a visible citation?
No. Being selected as a grounding source does not mean being linked as a citation, or being named in the answer. They are three distinct events.
Which tool shows the Selection Rate?
No engine displays it. Some GEO tools approximate it with retrieved vs cited sources: treat it as an estimate.
Terms related to Selection Rate
Further reading on Selection Rate
- Visibility in AI search: are tracking tools misleading us?
- Best GEO tools 2026: top 10 free and paid
- ← Back to the GEO glossary (126 terms)
Sources for the Selection Rate definition
- Dejan, “SRO & Grounding Snippets” (blog)
- Dejan, “Primary Bias” (blog, 2026)