Measuring AI visibility: What the numbers really say.
Mention, recommendation and source link answer different questions. A good AI audit will show you these differences and reveal the answers.
What is AI visibility?
AI visibility describes whether and how your company appears in the AI answers examined. It applies to the selected questions, languages, location specifications and the measurement time. It is not a market share or a measurement of all users.
The three most important key figures
| Key figure | What counts? | What it does not prove |
|---|---|---|
| Mentioned | Your company name appears in the answer. | A mention does not have to be a positive recommendation. |
| Recommended | Your company will be suggested as a suitable option. | A recommendation is not an order or a quality check. |
| Website as source | A page from your domain is linked as a source. | The link does not prove a visit and does not automatically explain the recommendation. |
A simple calculation example
Assume that ten answers to open customer questions can be evaluated: four mention your company, two recommend it and one links to your website. This results in 40% mentions, 20% recommendations and 10% references. These categories can overlap; you cannot add them to a total value.
In our report, the overall score corresponds to the recommendation rate. It is not an average of the three values. Answers to questions that already mention your brand appear separately. Technically failed or unclear evaluations are not simply counted as “not found”.
This makes the comparison of competitors fair
Compare companies in the same answers using the same counting rule. This means that a company that is mentioned five times in a single answer will not be visible in five different answers. In the event of a tie, both companies should receive the same rank.
The report shows the competitors appearing in the answers. This is not a complete list of your market. Also check whether a company mentioned actually offers the same service and is active in your target area.
Why repetition and good questions are important
Three repetitions may show that answers vary. They do not yet provide a high level of statistical certainty. Many almost identical questions can also create the impression of accuracy, even though only a narrow part of your offer is examined.
Cover different needs: location, specific services and selection criteria. Keep questions and settings as consistent as possible in later comparisons. Adding questions may change the results even if your website stays the same.
Compare languages and countries correctly
A German-language query with the location specified in Switzerland and a French query with the location specified in France are different measurement conditions. The percentage values refer to the answers from this group, not to the population or all search queries in the country.
Translations can shift nuances of meaning. You should therefore check the questions and group sizes actually used in the report. A location requirement in the API is not a confirmation that a real user from that country would respond the same way.
A comprehensible analysis can be checked
Retain original questions, answers, sources, model information and measurement time. This way you can check unusual results and compare later measurements. AI Brand Check provides the response details and a CSV export.