Evidence

Evidence includes

• Canonical prompt
• Exact prompt sent
• AI system
• Raw AI response
• Extracted observations
• Citations returned by the AI system
• Validation result
• Exclusion reason where applicable
• Provenance and methodology versions

Response list

Filter by prompt, AI system, validity, brand mentioned or not, recommendation, sentiment, rank, citation presence, Intelligence group, verification state, and timestamp.

Raw response and extraction

Raw AI responses are retained as source evidence under the applicable permission and retention policy. Extracted observations are structured interpretations. A corrected extraction creates a new extraction version; it must not overwrite the original raw response.

Citation states

• Citation supplied by the AI system
• Citation reachable
• Citation verified to support the associated claim
• Citation failed verification
• Citation not verified

An AI-system-provided URL is not proof that the source supports the generated claim.

Invalid responses and permissions

Invalid responses remain available for audit where permissions allow, but they are excluded from analytical metrics.

Report-summary access, evidence-metadata access, raw-response access, sensitive leadership evidence, and evidence export may use different permissions. Public reports must not expose raw responses or sensitive evidence by default.

• Understanding Report Scope
• Reliability, Evidence, and Interpretation