The questions never name a brand, so a brand that appears was volunteered by the AI. Questions are written the way a buyer or shopper would ask, across topics such as reliability, value, customer service, safety and ethics, and checked so that none points to a tracked brand.
- Engines. Each question is asked once to each engine the study lists (ChatGPT, Google AI Mode and Google AI Overview, plus Gemini in some studies), with live web search, from the study's market and in its language.
- Visibility. An answer counts once for a brand if it names it anywhere in the text. Position is ignored, and a link alone does not count.
- Ranges. Each figure has a 95% Wilson interval. When two brands' ranges overlap, treat them as tied.
AI answers vary from run to run, so each study is a dated snapshot rather than a permanent ranking. It shows whether a brand appeared and which sources were cited, not how favourably it was described.