How to measure AI share of voice
AI share of voice is your proportion of all brand mentions across a set of answers. The number is only as honest as its denominator: count only the competitors you chose to track and your share rises every time you stop tracking someone, which turns a metric into a dial.
What is the correct denominator?
Every brand mention detected, including brands you have not added. Tracked brands plus an Others bucket should sum to 100%, rather than your tracked set summing to 100% on its own.
The test is simple. Remove a competitor from your tracked list and see whether your share of voice rises. If it does, the metric is measuring your configuration rather than the market.
Keeping removed competitors in the denominator is the honest choice even though it makes your number lower. The answer genuinely did recommend them, and pretending otherwise inflates your score exactly when you stop paying attention.
How does share of voice differ from visibility?
Visibility is how often you appear. Share of voice is how crowded the answer was when you did. A brand can hold high visibility and low share by being named last in a list of eight.
What good looks like
Reported benchmark ranges. Treat them as orientation rather than targets, because they vary enough by category that your own trend is worth more.
| Position | Typical AI share of voice |
|---|---|
| Established player | 10 to 15% |
| Category leader | 25 to 40% |
Questions to ask a vendor about their share-of-voice number
If it is only tracked competitors, the number is gameable by editing a list.
Counting every mention rewards repetition, so a single answer naming one brand three times distorts the share.
An unpublished engine weight is an unfalsifiable number.
If the verbatim text is not retrievable, the figure cannot be audited.
Common questions
No. Recomputing history when you edit a list makes trends incomparable across time, which defeats the purpose of tracking one.
Covers how AI visibility is calculated and where the numbers mislead.