We document every data source, every validation step, and every place we're uncertain. If our data is going into compliance workflows and warranty decisions, you need to know exactly where it comes from.
We don't use AI training data for safety-critical fields. Every recall, every serial decode, every manual URL comes from a primary source we can point to.
Brand identification — the foundation of every lookup — goes through five independent validation layers before we return a result.
| Level | What it means | When to act on it |
|---|---|---|
| HIGH | Sourced from an authoritative government or OEM database. Cross-validated by at least 3 independent agents. | Safe for compliance, warranty decisions, and audit-required workflows. |
| MEDIUM | Sourced from one authoritative source. Validated by 1–2 agents. Minor uncertainty in one or more fields. | Appropriate for operational decisions. Review manually for high-stakes compliance use. |
| LOW | Inferred or partially matched. May be based on pattern matching or AI disambiguation with limited cross-validation. | Flag for human review before using in any compliance or financial context. |
null | We could not determine this field. The sources we checked returned no usable result. | Do not infer. Surface to the user for manual entry or alternative lookup. |