AI coverage and collection methods

Model API observations

The current collection code supports OpenAI (shown as ChatGPT), Anthropic Claude, Google Gemini, Perplexity, xAI Grok, Mistral, and DeepSeek. Availability depends on your workspace configuration, provider access, and successful runs.

These are model API responses. They do not reproduce every answer in the corresponding consumer app. Search tools, model versions, location, conversation history, and personalization can change the result.

Copilot-labelled results use an OpenAI proxy

The current Copilot collector calls OpenAI. It does not observe Microsoft Copilot. Treat results labelled “Copilot (OpenAI proxy)” as proxy observations, not evidence of visibility in Microsoft Copilot.

Google surfaces are separate

Gemini model responses are not Google AI Overviews or AI Mode results. Direct collection of those search surfaces is not included in the coverage described here. AI Overview records must not be counted as Gemini evidence.

Check the date and sample

A configured schedule does not guarantee a completed result. Check the latest successful run, prompts, model, market, and number of answers behind a score. Missing runs are missing data, not proof that a brand was absent. A single response is one observation, and scores are not estimates of all users’ experiences.