Methodology

The pipeline

Each requirement on this site moves through a fixed sequence before it is published. The steps are designed so that a human, not an automated process, makes the final call on what appears in public.

1. Discover candidate sources. For a given scenario and state, we identify the candidate primary sources that could govern it: the relevant federal standards plus the applicable state statutes, administrative codes, and DOT standards.
2. Approve the sources. A person reviews the candidate list and approves the sources that actually apply before any extraction begins. Nothing is extracted from a source a human has not approved.
3. Extract structured requirements. From the approved sources, we pull discrete requirements one at a time. Each is captured with a verbatim quote, the exact section, and the source URL, and is written as a draft.
4. Review every draft. A person reads each draft against the quoted source and publishes only what checks out. Drafts that do not match the source, or that are ambiguous, stay as drafts and are corrected or dropped.
5. Date and re-verify. Each published record is stamped with a last-verified date and re-checked over time, so entries stay current as editions and state manuals change.
6. Fold in corrections. Reports submitted through the Report an error control on a requirement page feed back into review, so real-world corrections improve the record.

What this guarantees

Only published, human-verified records appear on public pages. Drafts, candidate sources, and unverified extractions are never exposed to readers. If you see a requirement on this site, a person has read it against its source and decided it holds up.

For the reader-facing version of the trust story, see how we verify. To suggest a source, a scenario, or a correction, use contact.