Windtunnel method
Mention rate
Mention rate is the share of unbranded answers that name the brand unprompted. It is computed over eligible stored answers and carries a Wilson 95% confidence interval.
Counting only what the question did not plant.
The denominator is eligible unbranded samples: stored answers to brand-free questions like “What should I look for in an action camera?”. The numerator is the share of those answers that name the brand. A mention a branded prompt planted does not count: if the question names you, your presence is not visibility, it is the prompt’s structure.
Because answers are sampled repeatedly, the rate is a proportion over many stored answers, and it ships with a Wilson 95% confidence interval so the sample size is never hidden.
A public example.
The Insta360 audit in the published audit example reports a 94.9% mention rate (Wilson 95% CI 89.3–97.7, n=118, stored run from 19 Jul 2026, OpenAI + DeepSeek, ungrounded). The metric heads the Presence pillar of every Windtunnel audit report.
Three things a mention rate is not.
- Not a recommendation. Being named is not being ranked or advised; the recommendation family of metrics answers that separately.
- Not a ranking prediction. A rate describes stored answers on a tested date. It does not predict what a model will say tomorrow, and it is never a promise.
- Not a trend. One audit is a snapshot. Movement over time comes from a re-audit with the same prompts and method, a quarter later.
Labels travel with numbers: figures traced to stored answers from real model routes are MEASURED; Message Lift results are SIMULATED and comparative only; anything below the n≥30 gate or drawn from a single prompt cell is DIRECTIONAL. No figure on this site ships without its label, sample size, and date.