How we measure visibility
The check asks dozens of questions in Hebrew across two answer engines, reads the answers, and extracts from them who was mentioned and in what position. This page explains exactly how — including what the measurement cannot tell you.
What happens in a check
We pick questions a customer actually asks
Every industry has a bank of Hebrew questions — not translated English queries, but the phrasing people type into a chat box. The free check takes the highest-weighted questions from it.
We ask the engines
Every question goes to ChatGPT and to Gemini. We do not steer the model and we do not ask it for a list — the question is asked the way a customer would ask it.
We read the answers and pull the names out of them
A second pass over every answer extracts which businesses were mentioned, in what order, and which sources were cited. Order matters: being first in the list is not the same as being fifth.
We compute three metrics
Mention rate, average position and share of voice against competitors. All three are computed on discovery questions only.
Why only discovery questions count
A discovery question is a question that does not name the business. Only that kind of question can measure whether you get found. Ask "what do you know about the Cohen firm?" and the model will mention Cohen — it was asked to. That measures reputation and not visibility, and mixing the two is what inflates a score until it means nothing.
When we refuse to give a score
If every question we managed to ask named the business, there is nothing to compute visibility from. In that case the report says "we could not measure" instead of showing a zero. A zero and "cannot be measured" are two entirely different conclusions, and we do not show the first when the truth is the second.
Why two engines and not three
The check runs against ChatGPT and Gemini, the two engines consumers in Israel actually use. An adapter for a third model already exists in our code, but adding it at equal weight would give it a third of the score despite a marginal share of use — and that is a problem of measurement validity, not of cost. Per-engine weighting is a precondition for adding it.
What this measurement does not say
We measure through the providers' APIs, not through the app. A person talking to ChatGPT has a history, personal preferences and sometimes live browsing, so the answer they get can differ from the one we see. Answers also vary from one run to the next — that is the nature of a model, not a fault.
What the measurement does give is a consistent basis for comparison: the same questions, the same models, over time and against competitors. That is enough to see a trend and to know whether something worked — and it is not the same as what a particular customer saw on their screen. Anyone who promises you the second is selling you a certainty they do not have.
Google itself warns in its own guide against third-party tools that promise ranking success or claim to use its “internal” metrics. We are a third-party tool, and our answer is what is written above: we ask ChatGPT and Gemini the questions a customer asks in Hebrew, and we read what they answer in public. We have no access to any provider's internal metrics, and we do not promise rankings — not in Google and not in any other engine. Google's AI optimization guide
Want to see this for your own business?
The check is free and takes about a minute.
Run a check