You have probably done it already: opened ChatGPT and asked "which companies are good at [your field]?" to see whether you are on the list. A logical reflex, and as a measuring instrument almost worthless. Not because the answer is wrong, but because one answer says nothing.
The same model, three answers
AI assistants gamble. Literally: there is deliberate randomness in how they choose their next word, because otherwise all answers would be wooden and identical. Ask the same question three times and you get three different lists. In our measurements it constantly happens that a brand is mentioned in draws one and three and not in draw two.
Whoever concludes from one answer "we are in it!" or worse, "we have dropped out!", is steering on noise. In our own measurement history we saw a brand go from 56% to 22% in one day, same questions, same settings, and the next measurement it was simply back. That was not a drop; that was chance.
And you forget two thirds of the field
ChatGPT is not the only assistant your customers ask their questions. Gemini is in every Google environment, Claude is growing fast in business settings. And those three notably disagree with each other: in our comparative measurement the same brand appeared in 36% of the Claude answers, 5% in ChatGPT and 1% in Gemini.
Whoever looks only in ChatGPT and finds nothing there may be market leader in Claude, and vice versa. Without measuring all three you simply do not know where you stand.
How to measure it properly
Reliable measuring is not magic, it is discipline:
1. Ask the questions customers really ask. Not "what do you think of [brand]" but "which agency should I choose for a corporate video?", the question when nothing has been chosen yet. That is the moment when visibility is worth money.
2. Ask every question multiple times. Three times is the minimum on which you can say anything at all; more is better. And store every draw separately, so you can see how stable the picture is.
3. Reckon with a confidence interval. "Mentioned in 2 of 3 answers" is, with 95% certainty, somewhere between 21% and 94%. That range feels unsatisfying and is exactly what three measurements are worth. Only after weeks of measuring does the band get narrow enough to see trends.
4. Repeat it weekly. Visibility is not a photo but a film. The question is not only "am I in it" but "is it moving, and demonstrably so?"
The shortest route
This can be done by hand: three assistants, ten questions, three times a week, tallying in a spreadsheet. That is keeping track of ninety chats a week. Or you have it measured. Geony does exactly the above, every week, with an interval on every figure and every source traceable to the real answer.