AI share-of-voice calculator
Is your AI share of voice real, or noise?
Enter how many AI answers you sampled and how many mention each brand. The calculator shows each mention rate and share of voice with a 95% interval, and says when a gap is too close to call.
It runs in your browser: nothing you enter is sent anywhere. Free, no sign-up.
Your numbers
Questions × engines × runs. 60 questions on 2 engines, run once, is 120.
Paste a CSV instead
Your brand first, then each rival, as brand,mentions. A header line is fine, and so is a paste from a spreadsheet.
Your result
It appears here as you type.
Enter the answers you sampled and how many of them mention each brand, and you’ll see each rate with its error bars.
Now see whether an engine names you.
The numbers above are only as good as the sample behind them. Trovance runs each question on each engine every week (Claude on a sample) and measures the noise from questions it never acts on.
Worked example
Three brands, 120 answers
Northwind, Contoso and Fabrikam, with 60 questions put to 2 engines, run once: 120 answers.
How to read it
- Contoso is ahead of Northwind by 12.5 points (43.3% against 30.8%), but their 95% intervals overlap, from 34.8% to 39.6%, so at 120 answers the calculator flags it as too close to call.
- Share of voice tells the same story, loosely: Northwind holds 31.4% of the 118 mentions, with an interval from 23.7% to 40.2%.
- A real change of about 16.7 points in Northwind’s mention rate would show up in about four comparisons out of five at 120 answers a period; smaller real changes are easy to miss.
Too close to call at this sample size. Northwind’s 95% interval overlaps Contoso’s.
Mention rate
Answers that mention the brand, out of 120 sampled.
- NorthwindYou30.8%95% interval 23.3% to 39.6%37 of 120 answers
- Contoso43.3%95% interval 34.8% to 52.3%52 of 120 answers
- Fabrikam24.2%95% interval 17.4% to 32.6%29 of 120 answers
Share of voice (approximate)
- NorthwindYou31.4%95% interval 23.7% to 40.2%37 of 118 mentions
- Contoso44.1%95% interval 35.4% to 53.1%52 of 118 mentions
- Fabrikam24.6%95% interval 17.7% to 33.1%29 of 118 mentions
The smallest change you could see
16.7 points
If you sample 120 answers in each period, a real change of about this size in a mention rate of 30.8% would show up in about four comparisons out of five (95% confidence, 80% power). Smaller real changes are easy to miss.
Run more questions to see smaller changes.
Method
How the numbers work
Counts go in, a few formulas turn them into the numbers above. Here is what each one means, and where it stops being reliable.
Mention rate is the headline
It’s the share of your sampled answers that name the brand: if 40 of 100 do, the rate is 40%. A brand counts once per answer, however many times that answer names it, and one answer can name several brands, so every brand’s counts together can add up to more than the answers sampled.
The IAB defines it the same way: responses that contain a brand mention, divided by all responses.
The interval is the honest part
Ask the same questions again, or different ones, and the rate moves. The 95% interval is the range of rates your sample can’t rule out, and the width of the whisker is how far to trust the bar.
It’s a Wilson interval, which stays between 0% and 100% and holds up at small samples and at rates near either end, where the textbook formula doesn’t.
It measures the luck of the draw and nothing else. It can’t see bias in which questions you chose, and it treats every answer as independent, which repeats of one question on one engine aren’t quite. With many repeats, the true uncertainty is wider than shown.
Share of voice is an approximation
Share of voice divides a brand’s mentions by the mentions of every brand you entered, so it’s a share of that set, not of the whole market. Add or drop a rival and every share moves. The IAB defines it the same way, and asks providers to say which brands make up the set.
Its interval treats the total number of mentions as the sample. That’s a stretch, because several mentions can come from one answer and aren’t independent of each other. Read it as a rough guide to the split, and track mention rate.
Why fewer than 50 answers is exploratory
With 40 answers, one more mention moves a rate by 2.5 points, and the interval around it is wide. Below 50, read the bars as directions to check, not measurements.
The IAB’s measurement framework calls a program with fewer than 50 queries exploratory rather than directional. This calculator applies that floor to answers, which is looser: 50 answers can come from far fewer than 50 questions, so count your questions as well.
What “too close to call” means
When your mention-rate interval overlaps the top rival’s, the flag says too close to call. Overlap is a cautious test: this check can’t separate you, but two intervals can overlap a little and the gap can still be real.
Treat the flag as a reason to sample more, not a verdict either way.
The smallest change you could see
Compare two periods with N answers each. A real change in your mention rate of about 2.80 × √(2p(1 − p)/N), where p is your rate, would show up in about four comparisons out of five, at 95% confidence and 80% power. Smaller real changes are easy to miss.
Run more questions to see smaller changes.
Sources: IAB, Measuring Visibility in the AI Era · read October 9, 2026; Statistical Science, Interval Estimation for a Binomial Proportion · read October 9, 2026
The growth engine for the agentic web.
Get understood. Get cited. Get chosen.
Text Trovance
Or, in iMessage, send this note to +1 (839) 205-9508:
Hi Trovance, show me how AI sees my company.
You choose to send it. Trovance replies, and sends the updates you ask for, over iMessage through our messaging provider. Reply STOP anytime. Tapping or sending never bills you.