Key point. AI share of voice is the share of an engine's answers that name your brand. On 60 prompts, a 30% share has a 95% range of about 20% to 43%, so a 10-point lead over a competitor can be noise. This calculator shows the range and says when a gap is real.
AI share of voice calculator
Free calculator · runs in your browser
AI share of voice, with its margin of error
Enter how many prompts you checked in one AI engine and how many of the answers mentioned each brand. You get each brand's mention rate with a 95% range, and whether you are really ahead of or behind your nearest competitor.
Brands and the number of answers that mention each First row is your brand.
| Brand | Mentioned in | Mention rate (95% range) | Range | Share of mentions |
|---|
Share of voice here is the mention rate: the share of answers that name the brand. The share of mentions (your mentions divided by all brand mentions) is shown as a point only. Several brands can appear in one answer, so those mentions are not independent and a simple range for them would be too narrow.
How the ranges are calculated
Each mention rate uses the Wilson score interval at 95% confidence, with the prompt as the unit. The comparison with your nearest competitor uses Newcombe's method for the difference between two rates and treats the two rates as independent. When both brands were measured on the same answers, the true range can be somewhat wider or narrower. Never average rates across engines: each engine answers a different population of questions in its own way.
Built by Elementera. Lenz measures mention rates per engine for you.
What is AI share of voice?
We use the mention rate: of the prompts you checked in one engine, the share whose answer names the brand. It works with counts from any tool, including a spreadsheet you filled in by hand. Enter the number of prompts and, for each brand, the number of answers that mention it.
Why does it need a margin of error?
Every rate here is an estimate from a sample of prompts. The calculator gives each brand a 95% Wilson score range and compares you with the competitor whose rate is closest to yours. If the range on the gap includes zero, the two cannot be told apart at this sample size, and the calculator estimates how many prompts would separate them if the rates held.
Why one engine at a time?
ChatGPT, Gemini, Perplexity and Copilot answer the same question differently. An average across them describes no engine and hides where you are strong or weak. Run the calculator once per engine.
What about share of mentions?
Share of mentions (your mentions divided by all brand mentions) is shown as a point only. Several brands often appear in one answer, so those mentions are not independent, and a simple range on them would be too narrow.
FAQ
Should I count runs or prompts?
Prompts. Repeat runs of one prompt are not independent answers. If you ran each prompt several times, use one answer per prompt here.
How many prompts do I need?
It depends on your rate and the gap you want to see. The sample size calculator answers that directly.
Is it ahead if my range does not overlap theirs?
Yes, non-overlapping ranges mean a real gap. Overlapping ranges can still hide a real gap, which is why the calculator tests the gap itself.
Work with us
Want it done for you? Book a call about managed AEO services. Want to see where you stand first? Get your free AI visibility report from Lenz.