Every score can be checked. This is exactly what we do, and what we don't.
We ask the AI the questions a customer in your town asks, in your language. Without your name ('What is the best hairdresser in Leeds?') and with your name ('What do you know about Salon X?'). The free check asks 5 questions without your name and 1 with it. The report asks 16 without and 4 with. The templates are fixed per industry, so your score is comparable to a competitor's.
We query the AI models through their official APIs with web search enabled: ChatGPT (OpenAI), Gemini (Google) and Perplexity. An API answer is not identical to what you see in the app; we calibrate our model choice against the real apps regularly and name the model in every report.
AI answers vary. That is why we ask every question several times (2 times in the free check, 3 in the report) and show a margin. A score of 46 ±5 means: on repetition we saw between 41 and 51.
We recognise your business by name (with variants), domain name and phone number. A second, independent parser reads every answer and records which businesses are named and in what order.
Recommended (50%): in what share of the unnamed answers you are named, weighted by position. Known (15%): does the AI know your business when asked. Correct (20%): are phone, address and website right in the answers. Reputation (15%): the tone about your business, and your Google rating when known.
We don't scrape the apps (against their terms). We don't promise a position. We don't sell 'AI optimisation' without measuring before and after.