What is mention rate?

Mention rate is answers that name your brand divided by answers sampled. The link does not matter. A plain-text mention counts the same as a linked one.

Two details keep the number honest:

• Count answers that returned. A request that errored or timed out is missing data. It is not an answer without your brand.
• Fix the denominator in advance. Decide the prompt set, platforms and settings before you run them, not after.

The two diverge often. An engine can name you from its training data, or from a third-party page, with no link to you at all. Citations are also a noisy guide to influence. Mechanistic research (arXiv 2606.28358) repaired over 90% of missed citations by adjusting specific attention heads, without changing the sources.

You pick a prompt set, run each prompt on each platform, and record whether the brand is named. Then you divide.

The prompt set does most of the work:

• Branded prompts ("what does [your brand] cost?") name you almost by definition. They test accuracy, not visibility.
• Unbranded prompts ("best dental clinic in Austin") test whether you make the shortlist at all. This is usually the number a marketing lead cares about.
• Wording, location and account state shift results too. OpenAI says ChatGPT search rewrites your question into one or more targeted queries, and may use saved memories when doing so.

That rewriting means your prompts are a proxy for queries you cannot see. Retrieval is also discretionary: some answers use live search and some come from the model's training data alone. Any mention rate averages both.

The practical move: publish the prompt list with the number, run logged out with memory off, and keep the set stable over time. A reproducible prompt tracking method covers the setup. Remember that any sample is a sample. In Otterly's 2026 experiment, Microsoft's own report counted 42,817 citations for one test site while Otterly's prompt tracker logged 3,934, though the time windows may not match.

An unbranded mention rate tells you whether AI answers put you in front of buyers who have not heard of you yet. That is the answer that shapes a shortlist.

When engines draw on other sites instead of yours, being named on those sites is associated with being named in the answer. No effect size is promised. Third-party mentions and AI answers explains that route in depth.

Compare yourself with brands of similar size. Citation concentration is steep: a study of AI search (arXiv 2507.05301) found the top 20 news sources took 67.3% of all citations for OpenAI models. A small brand benchmarked against a giant mostly measures brand size.

For a first reading, the free checker on this site shows whether you are named for up to three buyer questions. You can also check whether ChatGPT mentions your brand.

• "Any mention is a win." An answer can name you and get your price, product or ownership wrong. See how often AI search gets facts wrong.
• "Mention rate is my AI ranking." It is presence or absence. No engine publishes position data to check a rank against.
• "One number covers every platform." Google says, as of its current documentation, that AI Overviews and AI Mode may use different models and techniques. Track each surface on its own, for example AI Overviews for your brand queries.
• "Markup raises it." There is no credible evidence that schema, llms.txt or AI-friendly phrasing gets you named. What you control is access: crawlers allowed per bot, reachable past your firewall, and key facts in the HTML before JavaScript runs. Those remove blocks. They do not guarantee a mention.

• Citation rate: the share of answers that cite your URL as a source.
• Prompt set: the fixed list of questions behind any mention rate.
• Retrieval state: whether an answer used live search or training data only.
• Platform coverage: which engines you sample. See which tools track Claude alongside ChatGPT, Gemini and Perplexity.

What is a good mention rate? There is no published benchmark, so treat any universal target with suspicion. The useful comparison is your own rate over time on a fixed prompt set, and your rate against competitors of similar size on the same prompts. A change in prompts, platform or settings breaks the comparison, so record all three every time you measure.

Is mention rate the same as share of voice? No. Mention rate asks whether your brand appears in each answer. Share of voice usually compares how often you appear relative to competitors across the same answers. Both depend on the prompt set. You can have a high mention rate and a low share if every answer also names five rivals.

Why does my mention rate change between runs? AI answers are not fixed. Engines rewrite queries, decide per question whether to search the web, and update models. Location, account state and memory also change results. Run the same prompts several times, logged out, and report the spread rather than a single figure.

Can I raise my mention rate directly? Not directly, and nobody can promise it. You can remove blocks: let each search crawler in, make sure your firewall does not turn it away, and put key facts in server-rendered HTML. Being named on sites that engines already cite is associated with appearing in answers, without a guaranteed effect.