How your AI visibility score is calculated
The scanner asks real questions to multiple AI engines and counts what is in the answers. Here you see the measurement chain, the difference between the first check and the full report, and the limits of the score.
We use purchase-oriented questions about sector and region, in the report built on real search queries from your own search data.
Every chosen question goes word for word to ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews.
For every answered result we check whether the brand is mentioned and actively recommended.
The visibility score follows from those counted signals.
The measurement protocol, step by step
Every scan uses the same sequence. This makes visible which signals go into the score and what stays outside the measurement.
A search question without a brand name
We use purchase-oriented questions about sector and region, in the report built on real search queries from your own search data. The discovery questions never contain the brand name, so the measurement shows whether the business appears spontaneously; the report may also include one control question about the brand itself.
The same question to multiple engines
Every chosen question goes word for word to ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews. We measure the first four through the model; for Google AI Overviews we read the actual answer block in the search results.
Mention and recommendation counted separately
For every answered result we check whether the brand is mentioned and actively recommended. A question without an answer does not automatically count as zero.
One fixed calculation rule
The visibility score follows from those counted signals. The figure is calculated, not a loose judgement an AI model gives on its own.
The first check and the full report do not measure to the same depth
The first check is deliberately small. The full report uses more questions and engines and adds context you need to choose improvements.
| First check | Full report | |
|---|---|---|
| Questions and engines | Two real search questions on one AI engine. | You choose the questions, built from your own search data; per question you see before and after, across all available engines. |
| What you see | A limited first visibility score. | Mentions, recommendations, competitors and sources. |
| Technical context | Not intended as a full audit. | A technical check and improvement points with priority. |
| Access | No existing account needed. | The current price is on the pricing page. |
What the score does not claim
A clear figure remains a snapshot. The limitations therefore belong next to the method, not in small print at the bottom.
Day, model version and phrasing can produce a different answer. The score is not a fixed final figure.
The chosen question panel stays the same between measurements, so before and after remain comparable. The questions themselves are built on real search queries from your own search data.
Repeat measurements are only meaningful when language, market and settings stay the same.
If an engine gives no answer, the calculation does not treat that as a negative judgement of the business.
What does and does not enter the question
The measurement needs context, but not data about your customers or employees.
Read how FlowCore handles dataSector, region and selected search questions provide the context. The brand name stays out of discovery questions so a spontaneous mention remains measurable.
The full report may also contain one question that checks the brand directly. It remains clearly identifiable as a control question.
Data about your customers or employees does not belong in model questions.
Data from Search Console, Analytics or other Google connections is not put into a question sent to an AI model.
First measure where you stand today
Start with the limited check without an existing account, or see on the pricing page what a full report contains.