01
Question set
Questions your audience actually asks. Buying intent, comparison and category questions mixed.
AI visibility explained
Being named is measurable, even without a click.
AI visibility describes whether and how often a brand turns up in generated answers. This page explains the measurement. The service behind it is the audit.
Definition
What gets measured.
AI visibility means the share of a system's answers in which a given brand, person or website gets named or linked, relative to a fixed set of questions.
The definition carries three conditions that are usually missing. It needs a fixed question set, otherwise nothing is comparable. It needs the system named, because values differ sharply. And it needs a date, because everything shifts continuously.
Without those three, a number for AI visibility cannot be interpreted. Anyone quoting a percentage without set, system and date has said nothing.
Method
How a measurement runs.
02
Pick systems
Two to four, chosen by audience. More raises the effort without adding insight.
03
Query repeatedly
Every question several times per system, split by region. One query is not a measurement.
04
Record
Named, linked, in which position, which source, who else. Plus date and screenshot.
Reading it
Telling noise from change.
These systems fluctuate. The same question can return different names twice in a row without anything having changed. Reading a trend into that is reading randomness.
A statement becomes solid across several measurement runs. We call something a change only once it holds over at least three of them.
Second trap: when a provider switches models, values move without any action from you. That is why every run gets documented, so that later on you can tell whether your visibility changed or the system did.
Common questions
About measurement.
Is there a visibility index like in SEO?
None that is recognised. Various vendors compute their own indices, which cannot be compared. We work with raw values per question and system instead of one aggregate number.
How many questions make a meaningful set?
Twenty at minimum as a rule of thumb. Below that the share swings too hard, because a single question already moves it by five points.
Is measuring ChatGPT enough?
For a first read yes, for a decision no. Values differ noticeably between systems, and you want to know where the gap sits.
What counts as a good value?
That cannot be answered in general, because it depends on the competition inside your question set. What is meaningful is the comparison with the providers named instead of you.
Can visibility be measured without a tool?
Yes, by hand. Ask questions, note answers, write down the date. It is tedious and it works, and for a small set it is a sensible way in.
Why screenshots?
Because answers cannot be reconstructed. Without evidence it later comes down to one claim against another, especially when a provider changes the presentation.
How often should measurement run?
Monthly is usually enough. More often rarely adds insight, because the variation between two queries is there regardless.
Does a mention without a link count?
Yes, and we record it separately. A mention without a link brings no click, works on awareness and reappears in follow-up questions.
What if we are described incorrectly?
That is a finding in its own right and more important than raw frequency. The cause almost always sits in a source that can be corrected.
How does this relate to the metrics?
AI visibility is the raw quantity the four disciplines build their metrics from. Which applies where sits under GEO metrics.
Next step
Let us find out where you stand in AI answers.
One conversation, 30 minutes, no sales pressure. You leave with an honest read on whether this is worth the effort for you.
