Key point. At a mention rate near 30%, about 320 prompts give a score within ±5 points, 19 times out of 20. Ten prompts give ±25 points. Use the calculator for your own rate and precision.
How many prompts do I need to measure AI visibility?
Free calculator · runs in your browser
How many prompts do I need?
An AI visibility score is a share of answers, and a share measured on a few prompts is a wide range, not a number. Tell us how precise you need it to be and get the number of prompts to track.
Prompts to track
-What a given number of prompts buys you
| Prompts | Your score would read | 95% range |
|---|
Why the prompt is the unit, not the run
Asking the same prompt five times does not give you five independent readings. The engine tends to answer one prompt in a similar way each time, so repeat runs mostly tell you about that one prompt. Different prompts tell you about your visibility. This calculator counts each prompt as one observation and adds credit for repeat runs only as far as you say they differ.
The ranges use the Wilson score interval at 95% confidence. Precision is half the width of that interval. Repeat runs are adjusted with the standard design effect, 1 + (runs - 1) × similarity.
Built by Elementera. Lenz tracks your mention rate across AI engines.
Why does a 10-prompt AI visibility score mislead?
An AI visibility score is a share: the answers that mention you, out of the answers you checked. A share measured on a handful of prompts is a wide range. At 30% on 10 prompts, the 95% range runs from about 11% to 60%. Two tools, or two months, can disagree by 20 points with nothing real behind it.
Why is the prompt the unit, not the run?
Asking one prompt five times does not give five independent answers. An engine tends to answer the same prompt in a similar way, so repeat runs mostly tell you about that prompt. New prompts tell you about your visibility. The calculator counts each prompt once and gives extra credit for repeat runs only as far as you say the repeats differ, using the standard design effect.
How is the number calculated?
The calculator finds the smallest sample whose 95% Wilson score interval, at your expected rate, is no wider than your precision on each side. If you do not know your rate, it plans for 50%, the worst case. The table shows what 10 to 1,000 prompts would buy at your settings.
FAQ
What precision should I aim for?
Pick the smallest change you would act on. If a 10-point move would change your plans, a precision of ±5 points or tighter lets you see it.
Does running each prompt three times help?
Only as much as the repeats differ. If repeats usually give the same answer, three runs of 100 prompts are worth little more than 100 prompts.
Can I combine prompts from ChatGPT, Gemini and Perplexity?
Measure each engine on its own. The engines answer differently, so a combined rate describes no engine in particular.
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.