# AI Reality Check

AI Reality Check is a deterministic continuity-assurance platform for AI workflows. It evaluates whether a changed AI output still preserves the approved user intent, required constraints, expected format, audience fit, and source/context requirements.

Canonical public domain: https://airealitycheck.co

Useful public references:

- Architecture summary: https://airealitycheck.co/architecture-summary.md
- Simulation and API testing guide: https://airealitycheck.co/simulation-testing-guide.md
- API quickstart page: https://airealitycheck.co/api-docs
- OpenAPI spec: https://airealitycheck.co/openapi/ai-reality-check-v1.yaml
- Human architecture page: https://airealitycheck.co/architecture
- Market signals and AI continuity FAQ: https://airealitycheck.co/market-signals
- Enterprise overview: https://airealitycheck.co/enterprise
- Scoring disclaimer: https://airealitycheck.co/scoring-disclaimer

Important boundaries:

- Scoring is deterministic and local/rule-based.
- The scoring engine does not call external AI APIs.
- The product does not use an LLM-as-judge for scoring.
- Enterprise API operations pass through a server-side authority boundary before scoring, report persistence, usage, audit, API-key operations, or unlocks.
- Tenant access, module access, credit deductions, report unlocks, audit access, and API-key access are never authorized by frontend state.
- Raw prompts and outputs are not stored by default for enterprise pilots; summary-only storage is the launch default.

How to evaluate:

Use the public architecture summary for conceptual review. Use the simulation testing guide for governed API smoke tests. Live API scoring requires an approved tenant and scoped API key.