OUHUDSearch Intelligence
Measuring AI visibility

Does an AI name your website — and for which question does it not?

Whether an AI system cites your page is controlled by nobody — not by us either. What can be controlled is the measurement: which question leads to you, which does not, and who stands there instead of you.

What the area does

Twenty answer surfaces, ten queried by us

ChatGPT, Perplexity, Gemini, Claude, Grok, Le Chat, DeepSeek, Qwen, Kimi and Llama have an interface — there we ask ourselves. Google AI Overviews, AI Mode, Bing/Copilot, Meta AI, Brave, DuckDuckGo, Yandex, Baidu, Naver and Amazon Rufus have none; they are observed through a SERP provider. Which surface is configured for you is stated in the interface — not as an advertising promise.

Every question repeatedly, otherwise there is no rate

Answer engines do not answer deterministically: the same question produces two different answers twice. A single hit therefore yields no rate. Every question is asked several times per surface, and the sample size stands next to the rate.

Cited and mentioned are two different things

“Cited” means your page was linked as a source. “Mentioned” means your company name appeared in the text, without a source. Both are counted separately, because one brings traffic and the other only awareness.

Roles and tone

The same question asked from different roles produces different rates — the run without a role remains as the reference value. In addition, the tone in which an answer speaks about your brand is evaluated. Without a substantiated judgement there is no “neutral” there, but nothing.

How you can tell

  • Every number carries its origin: queried by us means “measured”, observed through a third party means “reported”.
  • The denominator of every rate is the successful runs. Failed runs count neither above nor below the line — an unreachable surface is not a missing citation — and are stated anyway.
  • The wording of every question is stored unchanged. Change it and the trend compares two different things.
  • For every question it is stated which foreign domains were cited instead of yours.
  • There is deliberately no averaged “visibility score” across all surfaces: the systems access independently of one another, an average would say nothing.

Where it stops

  • ×A sample taken through the interface does not represent what a person sees in the consumer app.
  • ×No citation guarantee and no ranking guarantee. Answers change even without any change to your website.
  • ×Surfaces without a configured access stay off and appear as a gap — not as an estimate.
  • ×A single measurement says little. A direction only shows on the second measurement day.

Frequent questions

How many AI systems are really measured?

That depends on the configured accesses, not on an advertising promise. The registry knows twenty surfaces; which of them are configured today is stated on the home page in the measurement section and in the account per project. Ten surfaces have their own interface, ten are observed through a SERP provider — one access enables each of those for which the provider runs its own endpoint.

Why is the same question asked several times?

Because answer engines do not answer deterministically. The same question can cite your page once and not the next time, without anything having changed. A single run would therefore not be evidence; a rate from several runs is, as long as the sample size stands next to it.

What is the difference between a citation and a mention?

A citation links your page as the source of the answer — a visit can come from that. A mention only names you in the text, without a source and without a link. Both are counted separately because they are two different outcomes.