How many questions do you need to track AI visibility?

The number of questions you track decides how precise your numbers are. Here is how to choose it.

Written by the BotSay team. Last reviewed 23 September 2026.

1. The short answer
2. What a range tells you
3. More answers, a narrower range
4. What each plan gives you
5. All five engines, or one
6. How to choose your number
7. When to add more

The short answer

Enough that the figures you report come with a narrow range. Each question is read on five engines every week, so it adds five answers, and more answers give a narrower range. Every plan shows where you stand; a larger plan shows smaller moves, sooner.

What a range tells you

Mention rate is a share: of the answers we read, the share that named you. Counted from a handful of answers, it can swing on a single reply. Counted from a few hundred, it settles. So every BotSay rate comes with two things: the number of answers it is built on, and its likely range, the span the true figure almost certainly sits in.

More answers, a narrower range

Each week we ask ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews every question. So 60 questions give 300 answers a week, 150 give 750, and 400 give 2,000. The range narrows as the answers grow, though not in a straight line: four times the answers halves the range.

What each plan gives you

Here is one reading, a mention rate of 34.0%, at the size of each plan.

The likely range is how far the true figure may sit from the reading, either way. The smallest change we can call real is how far the rate must move in one week before the move is clear. It is about 1.4 times the likely range, because it compares two weeks at once. So two weeks' ranges can overlap a little and the change still be real.

The table shows each plan at its most precise. Answers to the same question on different engines tend to agree with each other, so real ranges run a little wider than these.

All five engines, or one

BotSay also shows each engine on its own. One engine gives a fifth of the answers, so its range is wider: at 34.0%, likely between 23 and 47 on Starter, 27 and 42 on Growth, and 30 and 39 on Scale. Use the five-engine figure for your headline, and each engine's figure, over a few weeks, to see where you are strongest.

How to choose your number

Start from the topics you want to win. Give each topic at least ten questions, written the way a buyer would ask. Ten questions give a topic 50 answers a week, a range of about ±12.7 points: enough to see where it stands. Read over four weeks, that is 200 answers and a range of about ±6.5. Add questions that name your competitors. That total is your plan, and we help you draft it from your site and your market.

When to add more

When a topic matters to you and its numbers still move from week to week, give it more questions. Reading a topic over several weeks together narrows its range in the same way, because it counts more answers.

The full method, including how we decide a change is real, is on How we measure. To check any AI visibility number, yours or a tool's, read Is my AI visibility score real?