Methodology version 1.0
How we measure AI visibility
- We generate realistic buyer questions that never name your brand or competitors, and filter out any that do.
- Each question is asked to several AI engines with web search on, several times, because answers vary between runs.
- An automated extractor lists the brands in each answer; quotes are verified against the answer text and positions are computed from the text.
- Mention rate is shown with a 95% Wilson confidence interval. The 0 to 100 score combines mention rate (40%), position (20%), share of voice (15%), citations of your own site (15%) and sentiment (10%); its range comes from bootstrap resampling of questions.
- Recommendations come from rules tied to evidence (metrics, sources, technical checks) and are ranked by impact, confidence and effort.
Limitations
- Results come from the official APIs of each AI provider with web search enabled. Consumer apps may differ because of personalisation, memory, location and ongoing experiments.
- AI answers change from run to run. We sample each question and show confidence ranges; small changes between audits may be noise.
- Visibility is measured on the specific questions in this audit. Different questions can give different results.
- Results reflect the date and time of the audit. No one can guarantee placement in AI answers.
- Brand detection is automated. Product names are trademarks of their owners; we are not affiliated with OpenAI, Anthropic, Google, Perplexity or Microsoft.