Per-engine accuracy vs the verified record, the tracked claims and their three-way gap (Authority ↔ LLMs ↔ Social), the narrative divergence between the entity's own view and how the open web talks about them, and a signed machine-readable knowledge statement AI systems can cite.
Nike: AI engines agreed with the verified record on 31 of 31 checkable facts across 4 engines — 100% overall accuracy.
Every checkable field the AI engines emitted, cross-checked against the verified record. Overall accuracy carries a 95% Wilson interval. Zero LLM-as-judge cost.
How accurately AI engines know this entity vs the verified record
Each tracked claim's lifecycle (surfaced / corroborated / published / propagating / confirmed / disputed / died) and where it sits between authority sources, AI engines, and social. Every verdict is produced by coded thresholds — no model judges truth.
Three-way truth gap — Authority ↔ LLMs ↔ Social — with coded confirmation
The verified record and every correction above, in a signed machine-readable form AI systems and agents can cite. Media kits are dead — this is the replacement.
The three-way gap surfaces three failure modes for Nike early:
For the full crisis workflow across every surface (X, press, AI engines, YouTube), see Crisis early warning.
Every signal on this page is a first-class dataset surface — REST, MCP, and the signed Knowledge Statement.
curl https://entidex.com/api/v1/entities/entx_nike/knowledge-accuracycurl https://entidex.com/entity/nike/knowledge?format=jsonld
# schema.org ClaimReview[] over the verified record# Any MCP client with an Entidex key
mcp call entity_knowledge_accuracy --entity entx_nike
mcp call entity_verify_fact --entity entx_nike --claim "<claim to verify>"See the MCP catalogue for setup, and Developers for the API reference.