Read the answer that produced the ranking.

A visibility score is only worth what it can show you. Every position in our charts traces back to a specific buyer question and the answer AI gave — the brands it recommended, the order it chose, and the words it used. Open the answer and check the number against it.

See this week's chartRead the docs
The buyer question, the model that answered it, and the response kept as it came back.

The buyer question, the model that answered it, and the response kept as it came back.

What you see

  • The buyer question exactly as we asked it, with the buyer type it was written for.
  • The answer AI returned, kept word for word rather than summarised.
  • Which brands were recommended and in what order, marked up inside the answer itself.
  • The model and version that answered, and the week it ran.
  • The same answers for every product named in them — including competitors.

How this is measured

We ask category buying questions, never questions about a specific brand. The answer is recorded in full and the recommendations are read out of it — a brand counts when AI recommends it, not when it merely appears in a sentence.

Because the prompts and the model version are held steady for the week, two answers from different weeks are comparable. That is what separates a trend from a single lucky reply.

Read the methodology

Common questions

  • What exactly do I see?

    The buyer-style question we asked, and the answer that came back — the brands AI recommended, the order it put them in, and the wording it used, kept as it was returned. Every position in the ranking traces back to one of these, so you can check the number against the answer that produced it.

  • Who writes the prompts?

    We do, from research into each category rather than from a keyword list. They are written the way a buyer would ask — by company size, what they are trying to do, and what the tool has to work with — and then held fixed week to week so a change in the answers is not a change in the question.

  • Does this show whether AI says good or bad things about my brand?

    No, and we would rather say so than imply otherwise. We never ask AI about your brand by name — we ask category buying questions and record who gets recommended. So there is no praise-or-criticism reading to give you. What you get instead is more decision-useful: whether you are recommended at all, where you are placed, and the words AI chose to describe you when it did.

  • Can I see the answers for a competitor?

    Yes. The prompts are the category's, not any one brand's, so the same answers show every product that was named. That is often the fastest way to see how AI frames a rival next to you, in its own words.

  • How is this different from asking ChatGPT myself?

    You can, and it is a good sanity check — the diagnostics check even walks you through doing it. The difference is repeatability: AI does not answer the same way twice, so a single try is a sample of one. We run a fixed prompt set every week on a fixed model version, which is what turns an anecdote into a trend.

Related features

All features