What each number means, and why you can rely on it.
AI visibility is how often, and how favourably, AI assistants name your brand when people ask the questions your buyers ask. This page explains how BotSay measures it. It is never behind a form.
Written by the BotSay team. Last reviewed 21 September 2026.
- The five measures, in one line each
- How sure we are of a number
- How we know a change is real
- Which questions count towards which number
- How each engine is asked
- Why every week
- How sentiment is scored
- What is outside our reach
The five measures, in one line each
Mention rate. Of the answers we read, the share that named you.
Share of voice. Of every brand mention in those answers, the share that was you.
Average rank. When you are named, how early. Lower is better.
Your pages cited. Of every link the answers gave, the share that pointed at your sites.
Sentiment. How favourably you are described, from 0 to 100, scored by a model.
How sure we are of a number
Every rate is counted from a known number of answers, and we print that count beside it. From the count we work out the likely range: the span the true figure almost certainly sits in. More answers give a narrower range, which is why we read every question on every engine each week. These are numbers you can stand behind.
How we know a change is real
In this example, last week's 27.8% was likely between 25 and 31, and this week's 34.0% is likely between 31 and 37. The two ranges do not overlap, so BotSay calls it what it is: a real change, up 6.2 points.
Which questions count towards which number
Questions that name your brand are reported on their own, because an answer to them names you almost by definition. Your headline figures come from the questions that do not name you, which is where winning a mention means something.
How each engine is asked
We ask ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews through their official interfaces, with web search switched on where the engine offers it, in English, for the market you choose. Every engine gets the same questions in the same week.
Why every week
A complete weekly reading means every Monday compares cleanly with the last: same questions, same engines, same method. That is what lets us say with confidence that you moved.
How sentiment is scored
A language model reads each sentence that names you and scores how favourably it describes you, from 0 to 100. We label the figure as estimated wherever it appears.
What is outside our reach
We measure what the engines answer when asked through their interfaces. A logged in person with a long chat history may see something different. We measure the questions you chose, so a good question set matters, and we help you write it.