Quick answer
Share of voice is your brand's mentions divided by total named-brand mentions, yours plus named competitors, across a sampled set of eligible answers. The worked example below uses illustrative numbers only, to show how the calculation and its caveats work, not as a measured result for any real brand.
The formula
Share of voice = your brand's mentions across the sampled answers, divided by the sum of every named brand's mentions in the same sample, including yours. Both counts come from the same fixed prompt set and the same run.
These numbers are illustrative
| Brand | Times named |
|---|---|
| Brand A (illustrative) | 18 |
| Brand B (illustrative) | 11 |
| Brand C (illustrative, you) | 9 |
| All other named brands combined | 12 |
In this constructed example, total named-brand mentions are 50. Brand C's share of voice is 9 divided by 50, or 18%, even though its raw mention rate across all 50 answers might be reported separately and look different depending on how many answers named no brand at all.
Reading the result correctly
- Share of voice is relative to named competitors, not to all possible answers. An answer that names nobody does not count toward anybody's share.
- Changing the competitor set changes the number without anything about your brand changing, so lock the set before comparing runs over time.
- Report it alongside the raw mention rate and the sample size, not as a single standalone figure.
How many sampled answers are enough?
Enough that the result does not swing wildly on a single rerun. There is no universal number; larger, well-defined prompt sets give more stable figures.
Should share of voice include unnamed, generic answers?
They affect the overall mention rate but are excluded from the share-of-voice denominator itself, since that denominator is specifically named-brand mentions.
Cite this page
BrandWater AI Research. (2026, 4 October 2026). Measuring share of voice across AI platforms: a worked example. https://brandwaterai.in/research/measuring-share-of-voice-across-ai-platforms-a-worked-example
How we work
Figures are dated and linked to their sources. Where none exist we say so. Read our methodology and AI transparency pages.