LLM Mention Tracking

Track a fixed prompt panel across ChatGPT and Gemini, preserving the underlying answers, citations, collection settings, and billing. Run inside the user's assistant or scheduler. UnifAPI supplies public data; this skill does not host a monitoring service.

Choose the right measurement

Connect with the unifapi skill and discover the current operation schemas and prices before collecting data.

• POST /geo/answers: executes one custom prompt in the ChatGPT or Gemini web interface. Returns separate observations.mentioned (brand-name aliases in answer text) and observations.cited (brand domain in answer sources). These observations support a fixed prompt panel.
• POST /geo/serp: samples Google AI Mode. It does not collect ChatGPT or Gemini. Its top-level is_target can match links as well as references; inspect answer references for citation evidence. best_rank is a result position, not a brand recommendation rank.
• POST /seo/serp with include_ai_overview: samples Google Search with AI Overviews. Keep it separate from AI Mode.
• POST /geo/mentions/search: searches an existing answer corpus; it does not execute each of your prompts. engine: google here means the Google AI Overviews corpus. The ChatGPT corpus is US/English only; that restriction does not describe live web-answer collection.
• POST /geo/mentions/cross-aggregated-metrics: returns counts for labeled target groups, not a ready-made share-of-voice score. Groups can overlap. Analyze these counts as a separate corpus study.
• POST /geo/mentions/history, /delta, /new-lost: provide historical corpus counts, signed changes, and newly observed/lost counts. These cannot reconstruct a customer's historical prompt panel. Coverage and collection changes can affect them.
• POST /geo/keywords/search-volume: estimates demand. Estimates are not actual prompt counts, traffic, or conversions. Freeze optional weights with the panel; report unweighted results alongside.

Workflow

Freeze the panel