We are not in the answers for our own category

We took the head question of our own market — what is the best AI brand monitoring tool? — and asked it of DeepSeek and Kimi, 3 times each. Six answers in total.

This is the question a buyer asks when they are already looking. It is the one worth being named in, and it is the one we built a product to be named in.

We were named in none of them. Not once, in six answers, on the question we exist to answer.

We are publishing that rather than a launch post, because a tool that measures whether brands are named in AI answers should tell you where it stands in them. If we only published results we liked, the numbers on this site would be marketing, and you would have no way to tell.

Same prompt, same model, two minutes apart — and a different answer. The models are not deterministic, and the difference is not noise around a fixed truth: a different set of brands comes back, in a different order.

So a tool that asks once and reports the result is not reporting a rate. It is reporting one draw, and it has no way to tell you whether the draw was typical. That is why we ask every question 3 times per engine and report a 95% interval next to the number — and why our reports say “not yet stable” instead of a decimal point they cannot support.

The uncomfortable version: the same criticism applies to the paragraph above. Six answers is a small sample. What we can say is that we were absent from all of them, which is a weaker statement than “we are never named” — and it is the statement the data supports.

If you are evaluating one of these products, including ours, three questions separate a measurement from a number:

• How many times is each question asked, and is that reported?
• Is there an interval on the figure, or a single value that looks precise?
• When an engine call fails, is it counted as “not mentioned” — or excluded from the denominator and shown to you?

A tool that answers all three can be checked. A tool that answers none of them is asking you to trust a decimal point.

The same thing we would tell a customer to do, which is the only honest answer: work out which pages are missing, publish them, and keep measuring. We will publish the follow-up to this note either way — including if the number has not moved.

Our own reports are the first thing we look at every morning. That is a strange position to be in, and it is the only one that keeps the product honest.

How the score is calculated, including the parts we know are weak.