---
title: "LLM Knowledge Accuracy — how accurately each AI engine (ChatGPT, Claude, Gemini, Perplexity, Grok) knows the entity vs the verified record: overall accuracy % (95% Wilson CI) plus a per-engine breakdown and the exact fields each engine gets wrong. The overall figure equals the Knowledge Statement accuracy. Coded cross-check over stored knowledge-probe evaluations — zero LLM cost."
method: GET
path: "/entities/{entityId}/knowledge-accuracy"
tags: ["Layer 1 — Profile"]
---

# LLM Knowledge Accuracy — how accurately each AI engine (ChatGPT, Claude, Gemini, Perplexity, Grok) knows the entity vs the verified record: overall accuracy % (95% Wilson CI) plus a per-engine breakdown and the exact fields each engine gets wrong. The overall figure equals the Knowledge Statement accuracy. Coded cross-check over stored knowledge-probe evaluations — zero LLM cost.

`GET /entities/{entityId}/knowledge-accuracy`

## Request

- Base URL: `https://entidex.com/api/v1`
- URL: `https://entidex.com/api/v1/entities/{entityId}/knowledge-accuracy`
- Auth: HTTP bearer

## Path parameters

- `entityId` string, required

## Response `200`

LLM Knowledge Accuracy, or { available: false } when the entity has no knowledge probes yet.

- union
  - LlmKnowledgeAccuracy: Per-engine accuracy of AI engines vs the verified record. The overall figure equals the Entity Knowledge Statement accuracy, promoted here to a first-class per-engine surface.
    - `entityId` string, required: Canonical entx_xxxx id.
    - `available` true, required
    - `computationVersion` string, required
    - `accuracyPct` integer, nullable, required: Overall accuracy across all engines — the headline %. Null when nothing decidable.
    - `evaluated` integer, required
    - `matches` integer, required
    - `mismatches` integer, required
    - `modelNulls` integer, required
    - `decidable` integer, required
    - `enginesEvaluated` integer, required: Engines that contributed at least one decidable check.
    - `confidence` ScoreConfidence, required: 95% Wilson confidence envelope under a Score (confidenceLayer.ts). All fields optional.
      - `interval95` number[]: [lower, upper] on the same 0..N scale as the score value.
      - `variance` number, nullable
      - `sampleSize` integer: Observations the score is computed over (enginesEvaluated for Decision Survival).
      - `probeRunIds` string[]: Contributing probe_run UUIDs (capped at 20).
      - `lastObservedAt` string, date-time
      - `version` string: Confidence-Layer scoring version.
    - `perEngine` EngineKnowledgeAccuracy[], required
      - `engine` string, required: Canonical model label, e.g. "ChatGPT".
      - `maker` string, required: Maker company, e.g. "OpenAI".
      - `makerHref` string, required: Maker entity page, e.g. "/entity/openai".
      - `evaluated` integer, required: All fields cross-checked for this engine.
      - `matches` integer, required
      - `mismatches` integer, required
      - `modelNulls` integer, required: Fields the engine made no claim about (coverage gap, not an accuracy miss).
      - `decidable` integer, required: matches + mismatches — the accuracy denominator.
      - `accuracyPct` integer, nullable, required: matches / decidable, or null when nothing decidable.
      - `corrections` KnowledgeAccuracyCorrection[], required
        - `field` string, required: Human-readable field label, e.g. "Headquarters".
        - `aiClaim` string, nullable, required: What the engine claimed (null for an ai_unaware gap).
        - `verifiedValue` string, nullable, required: The verified value the engine should hold.
        - `source` string, nullable, required: Provenance of the verified value, e.g. "Wikidata".
        - `kind` 'contradiction' | 'ai_unaware', required
    - `methodology` string, required
    - `provenance` object, required
      - `issuer` string, required
      - `methodologyUrl` string, uri, required
      - `note` string, required
  - LlmKnowledgeAccuracyUnavailable
    - `entityId` string, required
    - `available` false, required
    - `reason` 'no_knowledge_probes' | 'disabled', required

## Other responses

- `404`: Entity or resource not found.

---

[API](https://skmtc.dev/entidex/apis/entidex-api.md) · [All operations](https://skmtc.dev/entidex/apis/entidex-api/llms.txt) · [OpenAPI document](https://skmtc.dev/entidex/apis/entidex-api/revisions/86b92f21a09a?raw)