Glossary · Beginner
What Is Recommendation Rate? Definition and Example
Recommendation rate is the percentage of AI answers, across a defined prompt set, that actively recommend your brand as a good option for the user's need. It's stricter than mention rate: being listed in passing isn't enough. Because recommendations are closest to a referral, many teams treat this as their headline GEO metric.
How do you calculate it?
Recommendation rate = answers recommending you ÷ total answers scored
Illustrative: 30 prompts × 3 engines × 2 runs = 180 answers. If 27 recommend you, your recommendation rate is 15%.
What counts as a recommendation?
Agree a written rule before scoring, and apply it consistently.
| Answer wording | Recommendation? |
|---|---|
| "Brand X is a strong choice for small teams" | Yes |
| "Top picks: Brand X, Brand Y" | Yes |
| "Other tools include Brand X" | No, mention only |
| "Brand X is popular but expensive; consider Y" | No |
How should you segment it?
Report by engine, by intent (learn, compare, buy) and by product line or market. A flat overall rate can hide a strong position on learn prompts and a weak one on buy prompts, which matter more commercially.
Example
Illustrative. A brand has a 50% mention rate but a 12% recommendation rate. Reading the answers shows engines describe it as "an older option". The team updates its comparison page, refreshes review profiles and publishes current feature details, then tracks whether the gap narrows over three months.
Why it matters for GEO
Recommendation rate connects AI visibility to buying decisions more directly than mentions. Track it alongside mention rate and citation share; the gaps between them show what to work on.
Common pitfalls
- Counting a list of ten tools as ten recommendations; agree a rule on long lists.
- Changing the definition between months, which breaks the trend.
- Reporting a single run; recommendations vary between runs, so use several runs and look at trends.
Related
Questions
How is recommendation rate different from mention rate?
A mention only names you. A recommendation presents you as a good choice for the user's need.
What's a good recommendation rate?
There's no universal benchmark. Compare against your competitors on the same prompts and track your own trend.
Should branded prompts count?
Report them separately. Branded prompts inflate rates and are better used to check accuracy.