Many answers name several brands but only recommend a few. Recommendation rate isolates the cases where the engine positions the brand as a choice: 'consider X', a top-3 shortlist, the suggested option for a use case. It is computed from the same prompt portfolio runs as mention rate, with a stricter classification.
Using it in practice
- Define the classification rule before measuring: what counts as a recommendation versus a neutral mention must be consistent across runs.
- Record the position when answers rank options: being ninth on a list of ten is technically a recommendation but commercially weak.
- Note the reason the engine gives: those phrases reveal which evidence drives the recommendation and what to reinforce.
- Track sentiment for the rejections too: 'X is popular but expensive' is actionable intelligence.
Recommendation rate moves slower than mention rate because it depends on comparative evidence: reviews, case studies, third-party comparisons and the clarity of who the product is for. That makes it the best long-term indicator of an AEO program's commercial value.