How we measure
We ask AI systems the questions your customers ask, read each answer, and count whether your name is in it. This page explains what that number means and what it does not mean.
What we ask
From your site we propose questions a customer could ask, in your language and for your country. They do not contain your name. You check them first. In the paid scan you can edit them.
What we count
An answer counts as named when your name is in its text. We report it as a count per system: named in X of Y answers. We do not make a score, because a score hides what it is made of.
Why we repeat
AI systems answer differently each time. The free scan asks each question once, which is why it is called a first indication. The paid scan asks each question five times on six systems.
Evidence
Each number opens the answers it counts, with your name marked. Each statement in a finding links to raw data. An automatic check removes statements without such a link before the report is shown.
Measured, derived, general advice
Measured means we read it in the data. Derived means we conclude it from the data, and we can be wrong. General advice comes from primary sources such as official documentation, and its effect on AI answers is not measured. Each finding says which is which, and how sure we are.
When something fails
We retry and use backup routes. A run that still fails is shown as failed and stays out of the count. If fewer than half of the answers succeed, a free scan does not count.
What we cannot see
We see what these systems answered on the day of the scan, for the country and language you chose. We cannot see personal context in someone's own account. Perplexity is measured through its API, so it may differ from its website.