What Is Share of Voice in AI Search? – Definition and Measurement
Learn what share of voice means for AI answers, how to compute it from repeated runs, which denominators are valid, and how to read the result.
Share of voice in AI search is the proportion of brand mentions in generated answers that belong to you. It converts an absolute count into a competitive position, which is what a mention rate alone cannot express.
1. Why a Mention Rate Is Not Enough
Being named in 40% of answers means something different depending on whether two competitors or twenty are named in the rest.
Share of voice supplies the denominator: it measures presence relative to the field actually appearing in the same responses.
2. Computing It
share_of_voice = your mentions ÷ all brand mentions, across the same responses
The essential discipline is that the denominator comes from the responses themselves, not from a list you decided on in advance. Engines name brands you did not consider competitors, and that is signal rather than noise.
3. What to Count
| Choice | Effect |
|---|---|
| Any mention | Broadest; includes passing references |
| First mention only | Emphasises prominence |
| Weighted by position | Approximates visibility within the answer |
| Citations only | Stricter; measures being used as a source |
Pick one and keep it fixed. Mixing definitions between periods destroys comparability more thoroughly than any measurement error.
4. Sampling Correctly
Generated answers vary between identical runs, so share of voice must be computed across many runs, not one.
- Freeze the prompt set and its wording.
- Run each prompt several times per period.
- Pin location, language, and engine version where possible.
- Store full responses so metrics can be recomputed.
- Report the figure with its spread.
5. Segmenting
An aggregate share hides where you are strong and weak. Segment by prompt type — category questions, comparisons, problem-led questions, brand questions — because performance usually differs sharply between them.
A high share on brand questions and a low one on category questions is a specific, actionable finding that the aggregate conceals.
6. Reading the Result
- Rising share with a flat mention rate means competitors are being named less, not that you are named more.
- Falling share with a rising mention rate means the field is expanding.
- High share on few prompts may reflect a narrow niche rather than strength.
- Small differences are noise. Compare against the variance before drawing conclusions.
7. Pairing It With Accuracy
Share of voice counts mentions; it says nothing about whether the description is right. A high share built on outdated or incorrect characterisation is a problem wearing the appearance of success.
Extract the sentence around each mention and classify it, and report accuracy alongside share.
8. Summary
Share of voice measures your proportion of brand mentions in generated answers, computed across repeated runs of a frozen prompt set with the denominator taken from the responses themselves.
Fix the counting rule, segment by prompt type, compare movements against variance, and always report descriptive accuracy beside it — presence and correctness are different questions.