Methodology

What we do. We put real buyer-intent questions (generated for your category and country, without naming the brands) to AI engines, and we analyze in what order, how often and with what sentiment each brand appears in the answers.

How we query each engine. We use only each provider's official APIs with their web-search tool enabled (OpenAI, Google Gemini, Anthropic Claude, Perplexity, xAI) and authorized SERP providers for Google AI Mode / AI Overviews. We never automate the consumer interfaces. Every answer stores the engine, the exact model, the method and the date.

Limits, told plainly. An API-with-search answer is not an exact mirror of what a specific user sees in the consumer app (their memory, personalization and the provider's A/B tests all play a part). Independent studies show meaningful divergence between the two. That's why we say what we measure and how, and never promise guaranteed positions: the engines change without notice.

No made-up data. If an engine doesn't respond, it's marked as “no data”. Mention and sentiment analysis is based on the real text of each answer, which you can read in full in the report alongside the sources the engine cited.

The technical audit is different: it uses no AI engines. We check your robots.txt (access for bots like GPTBot, ClaudeBot or PerplexityBot), the presence of llms.txt, structured data (schema.org/JSON-LD) and whether your content is visible without running JavaScript. These are deterministic checks on what your server returns on the first response. The robots.txt heuristic detects full-site blocks; it does not reproduce the complete path-specific algorithm.