AI Engines/Anthropic
live monitoring

Claude AI Visibility Monitoring

GeoScript monitors how Claude responds to business and commercial queries, tracking mentions, recommendation rate, and response content across your prompt universe.

How GeoScript observes this engine

  • GeoScript sends deterministic prompt sets to the Anthropic Messages API (api.anthropic.com/v1/messages) via a live adapter.
  • Each call targets the production Anthropic endpoint with real credentials — not a simulation.
  • The raw HTTP response body is preserved before parsing. Claude's response is an array of typed content blocks; GeoScript concatenates all text-type blocks to form the observation text.
  • Observation latency, input/output token counts (including cache tokens when applicable), and actual model version are recorded.
  • Retries occur only on server-directed 429 or 5xx responses. Timeouts are terminal.
  • Prompts use consistent, temperature-controlled templates to support valid comparison across observation runs.

Signals available

  • Business mention detection — whether your business appears in Claude's response.
  • Recommendation rate — proportion of prompts where your business is mentioned or recommended.
  • Response text — the complete verbatim answer from Claude, preserved per observation.
  • URL extraction — HTTP/HTTPS URLs mentioned in the response text, extracted via pattern matching. Claude does not emit structured citation objects.
  • Historical tracking — results accumulate to show how Claude's responses change over time.
  • Observation metadata — prompt, engine, model version, latency, token costs, and timestamp.

What GeoScript does not measure

Explicit limitations (§127)

  • GeoScript cannot observe dwell time, user engagement, or behavior after Claude returns a response.
  • GeoScript does not have access to Anthropic's model internals, training data, or ranking parameters.
  • Claude does not emit structured citation objects. URL extraction is best-effort and may miss references or include false positives.
  • GeoScript observes real API responses — not user-session replays. Personalized or context-dependent responses a specific user sees may differ.
  • Observation results reflect a point in time. GeoScript cannot predict or guarantee future responses.
  • GeoScript does not measure any Anthropic product surfaces beyond the core chat/completion API.

Related

Other AI engines