Definition
Mention rate is the share of AI-generated responses, within a tracked set of prompts, that substantively mention your brand in the answer text. It is calculated as responses that mention the brand divided by all analyzed responses, usually per platform and per time period. A link is not required: an unlinked recommendation counts, because the reader still sees your name.
The unit is the response, and each response counts once per brand no matter how many times the name appears. If a brand is named in 55 of 200 ChatGPT responses in a month, its ChatGPT mention rate for that month is 27.5%. At that sample size the figure is accurate to roughly plus or minus six percentage points, so small month-to-month moves are often noise.
What counts as a mention matters as much as the formula. A mention means the answer discusses the brand: names it in prose, lists it as an option or includes it in a comparison table. A brand that appears only in a cited URL, a source title or a publisher name has not been mentioned, since the answer used its content without recommending it. Replies such as 'I don't have information about that company' are not mentions either. Promptwatch applies these rules when it counts one mention per brand per response, and the same rules keep mention rate comparable across tools.
Mention rate sits beside several related metrics:
- Citation rate counts responses that link to your domain as a source. The two move independently: content work drives citations, while reputation and third-party coverage drive mentions.
- Brand inclusion rate measures coverage at the prompt level (how many of your prompts include you at all), while mention rate works at the response level, which matters when each prompt is run many times.
- Ghost citation rate is the share of your mentions that carry no link to your site, the exposure that never shows up in referral analytics.
- AI share of voice is the competitive view: your mentions as a share of all brand mentions in the same responses.
- Answer position and visibility scores capture prominence, which mention rate ignores. Being the headline recommendation and being the ninth row of a table both count as one mention.
Because of that last point, mention rate works best as the first number in a diagnosis rather than the only one. A low rate is a coverage problem: models do not associate the brand with the category, or the platforms rely on sources that never discuss it. A high rate with a poor position means the brand is known but treated as an alternative. A high rate with negative sentiment is a reputation problem.
A few measurement habits prevent misleading numbers. Maintain a list of brand aliases, abbreviations and common misspellings, and check context for short or generic names that collide with ordinary words. Track product names separately when they appear in different kinds of prompts than the company name. Report mention rate per platform, because live-search products and API models without search draw on different evidence, and a blended rate hides the platform where you are losing. Our guide to tracking brand mentions in AI models covers setup in more detail, and how to get your brand mentioned by AI covers the work that raises the rate.
Examples of Mention Rate
- A project management tool is named in 55 of 200 ChatGPT responses in September, a 27.5% mention rate, against 41% for its main competitor on the same prompts.
- A team finds its mention rate on Perplexity is twice its citation rate, meaning half its appearances come with no link to its site, and starts reporting the ghost share alongside referral traffic.
- An analyst named 'Atlas Analytics' discovers its mention count is inflated by answers about maps and atlases, and adds a context check so only mentions in analytics prompts count.
- A brand with a 60% mention rate learns from position data that it is usually the fourth or fifth brand named, and shifts its goal from appearing more often to being recommended first.
