Two numbers, two questions
When AI assistants answer "who's the best plumber in Denver?", they usually return a short list. Two measures summarize what happened across many such answers:
- Mention rate — in what share of answers was the business named at all? If it was named in 6 of 40 answers, the mention rate is 15%.
- Share of voice — of all the times any business was named, what share went to this one? If the 40 answers named businesses 150 times in total, and the client got 6 of those, its share of voice is 4%.
Mention rate answers "do we show up?". Share of voice answers "how crowded is the list, and who owns it?".
How it's calculated
Each answer is checked for every business it names. Names are normalized so that small variations — "& Sons", "LLC", a city suffix — count as the same business. Each business gets one mention per answer that names it, however often it is repeated within that answer. Share of voice is that business's mentions divided by all business mentions across all answers. The same calculation runs for each competitor, which produces the competitor table.
How to read it
- High mention rate, low share of voice: the client shows up, but in long lists. Worth checking the position as well.
- Low mention rate, one dominant competitor: one business owns the answers. Compare its reviews, listings and service pages with the client's — that's usually where the gap is.
- Many competitors, each with a small share: the category is fragmented, and assistants disagree. In our sample report about 60 different roofers were named across 40 answers, and only 2 were named by all four assistants. That's an opening: no one owns the answer yet.
A worked example
Say a dental practice is checked with 10 questions across 4 assistants: 40 answers. It is named in 8 of them, so its mention rate is 20%. Across the 40 answers, the assistants named businesses 160 times in total. The practice's 8 mentions are 5% share of voice. The most-named competitor appears in 24 answers: a 60% mention rate and 15% share of voice. Next month the practice is named in 12 answers, but the total grows to 180 mentions because the assistants listed more practices. Mention rate rises to 30%, share of voice to about 7%. Both moved in the right direction, and the competitor table shows who lost ground.
Reported together, the two numbers stop a client from over-reading either one: a rising mention rate in a market where every list is getting longer is good news, but not as good as it looks on its own.
Per assistant, not just overall
An overall share of voice can hide big differences. One assistant may name a business in most answers while another never mentions it, often because they rely on different sources. Reporting mention rate per assistant, and the sources each one cites, shows where the work should go.
Limits worth stating to clients
- Share of voice depends on the questions asked. Ask different questions and the numbers change, so keep the question set stable from month to month.
- Answers vary between runs. Treat a single month as a snapshot and look at the trend.
- Answers collected through APIs can differ from what a person sees in the consumer apps, which also personalize by location and history.
- Automatic name extraction can occasionally miss or mis-parse a name, so very small differences between competitors aren't meaningful.
Where AnswerRoster fits
Every AnswerRoster report includes mention rate, per-assistant results, position when listed, the competitor share-of-voice table and cited sources, under your agency's brand. See the example walkthrough or pricing.