What mention rate means
Mention rate is the ground-level metric of AI visibility: out of all the AI answers generated for a tracked set of prompts, in what fraction does your brand get named at least once? If you track 50 category prompts across four engines — 200 answers per sweep — and your brand appears in 68 of them, your mention rate is 34%.
It is deliberately binary per answer: an answer either names the brand or it doesn't. Prominence, sentiment, and framing are measured separately; mention rate isolates the most basic question — are we in the conversation at all? — because that's the threshold everything else depends on.
The formula
mention rate = answers naming the brand ÷ total answers × 100
Three conventions keep the number honest:
- A fixed prompt panel. The denominator must be stable between measurements; change the questions and you've changed the metric. See prompts and the prompt matrix.
- Repeated sampling. LLM answers vary run to run — the same prompt can name you today and not tomorrow. Scheduled prompt tracking averages out that noise.
- Consistent mention detection. Deciding whether "the tool formerly known as X" counts as naming you is a judgment call, which is why serious trackers use an LLM-as-judge applying the same criteria every time.
Mention rate is usually sliced per engine (ChatGPT vs Perplexity can disagree sharply about the same brand) and per prompt group (branded questions vs unbranded category questions).
Mention rate vs share of voice
The two are easily conflated. Mention rate measures your own presence: the fraction of answers you appear in, regardless of who else appears. Share of voice measures competitive position: your slice of all brand mentions handed out. A brand's mention rate can hold steady while its share of voice erodes — engines still name it, but now name three rivals alongside. Tracking both catches that pattern early. Mention rate also differs from citation rate, which counts your domain being used as a source, not your brand being named in the answer text.
What a good mention rate looks like
On unbranded category prompts ("best X for Y"), category leaders often exceed 60–70% while challengers sit in single digits — the distribution is top-heavy because engines converge on a small consideration set. On branded prompts ("what is BrandName?"), anything below roughly 100% signals an entity problem worth investigating. As with all AI visibility metrics, the trend on a stable panel matters more than the absolute number, and the per-prompt detail — which questions you're absent from — is what turns the metric into a to-do list. It's also the primary input to any composite AI visibility score.