Skip to the content
searchboar.

AI search and visibility

Mention rate: how often does an AI answer name a brand?

Mention rate is the proportion of answers in a defined sample that mention a selected brand. Each answer is counted once even if it repeats the name. A meaningful rate requires a stated question set, assessment method and denominator.

Also known as: mention rate · AI mention rate · brand mention rate

At a glance

  • The numerator is answers mentioning the brand; the denominator is the answers assessed.
  • A mention is not necessarily positive, a recommendation or a link.
  • Question selection, ambiguous names and failed observations affect the result.
On this page
The same blue business symbol appears within the text of some answer cards.
Mention rate examines which answers in a defined sample name the brand.

What does the percentage count?

For a simple answer-level measure, divide the number of assessed answers that mention the brand by the total number of assessed answers, then multiply by 100.

Mention rate = answers mentioning the brand ÷ assessed answers × 100.

The unit is an answer, not each occurrence of a name. A provider may use a different metric definition, so check its documentation before comparing dashboards. The formula is a stated method, not a universal industry standard.

Example: six mentions in twenty answers

Imagine a fictional carpentry business in Szombathely. Its name appears in six of twenty successfully captured, assessable answers. The mention rate for that sample is 30%.

If one answer repeats the name four times, the numerator remains six. If two further requests failed, report the six mentions out of twenty assessed answers and the two failed observations separately. Failed retrievals do not establish that the brand was absent.

A percentage without sample size conceals how sensitive the result is. In a sample of twenty answers, one additional mention changes the rate by five percentage points.

How do you recognize a genuine brand mention?

A name such as “Sunshine” may refer to ordinary language, a different company or the intended business. Abbreviations, spelling variants and inflected forms also complicate matching.

Define which names count and inspect uncertain cases. Review a sample of both positive and negative classifications: checking only detected names can miss cases where the method failed to recognize a real mention.

The point is to establish that the answer refers to the business, rather than merely containing the same sequence of letters.

How does a mention differ from a citation or recommendation?

An AI citation identifies a source, often through a URL. An answer can name a business without linking to its website, or cite a page without displaying the brand name in its prose.

A brand recommendation is different again. “This company does not provide the service you need” contains a mention but is not a recommendation. A name appearing in a comparison or warning should not automatically be reported as positive visibility.

For these reasons, citation rate and mention rate should remain separate measures with their own definitions.

When are two measurements comparable?

Keep the question set, platform, language, location assumptions and assessment rules consistent. Branded questions naturally make the name more likely to appear than unbranded requests for a suitable provider. Mixing those groups can change the rate without any improvement in discoverability.

Semrush and Ahrefs document their own visibility methods. Read those definitions instead of assuming similarly named numbers are interchangeable.

Repeated observations can show how a controlled sample changes. They do not establish what every user sees, how well known the business is across the market or how many customers it acquires.

Sources

  1. Semrush: Prompt Tracking
  2. Ahrefs: AI visibility metrics
  3. Ahrefs: Setting up custom prompts for AI visibility tracking