Method
How we measure AI visibility
AI answers vary from one ask to the next, so a single screenshot proves nothing. This is how we turn them into a number you can trust and track.
1. Write the questions buyers actually ask
We write 50 questions for your market (10 for the free check): category questions ("best memory foam mattress UK"), use-case questions ("best mattress for side sleepers"), comparison questions ("Simba vs Emma") and price questions ("best mattress under £500"). You approve them before anything runs.
2. Ask every assistant, three times
Each question goes to ChatGPT, Google Gemini, Google AI Overviews, Google AI Mode, Microsoft Copilot, Claude, Perplexity, Grok, DeepSeek and Kimi. ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek and Kimi are asked through their official APIs with web search on, the way their consumer apps answer shopping questions. Google AI Overviews, Google AI Mode and Microsoft Copilot are read from the live search results page for your country. Every question is asked three times per assistant.
API answers are close to what app users see but not identical, because the apps add their own instructions and features. We label results as coming from APIs and live search pages, never as screenshots of the apps.
3. Score the answers
- AI visibility score (0–100): the percentage of successful answers that recommend your brand.
- Average position: where you appear in the list when you are named (1 = first).
- Top-3 rate: how often you're one of the first three brands named.
- Citation rate: how often the assistant links to your own website.
- Stability: how consistent your mentions are across the three runs.
Failed or empty answers (for example, when Google shows no AI Overview for a query) are recorded and excluded from scores, and the report says how many there were.
4. Map the sources
Every page each assistant cites is recorded. For every question where a competitor was named and you weren't, we list the sources that backed the competitor and never mention you. Those pages are usually the fastest route into the answers.
5. Turn gaps into fixes
Each fix is tied to evidence in the data: a missing page, a list you're not on, a fact the assistants get wrong. Fixes are ranked by likely impact and effort. A person reviews every report before it goes out.
What we don't do
We don't buy placements, write fake reviews, or promise rankings. Nobody can sell a guaranteed spot in an AI answer.