We check that every primary source we cite is still live, we keep an append-only record of every change we observe — and when we first saw it — and we invite you to report any error. Everything below is a live count over real records.
AI Law Tracker is a primary-source legal-information product, so the honest question a reader should ask is: how do you know it's right, and how would I catch you if it weren't? This page is our answer. Two automated systems run against our own data — one that re-checks whether every government source we link to is still live, and an append-only history that timestamps every change the moment we record it — and their output is published here, unedited. None of the numbers on this page are written by hand; they are live counts over real rows.
Source-health check
A scheduled job re-fetches every primary government URL we cite and every curated source link, then records a verdict for each. A source is only ever flagged as broken on unambiguous evidence — a hard 404/410, or a page that returns 200 but is really a “not found” husk. A source we simply couldn't reach (a bot or geo block) is marked unverified, never “broken”. A flag never edits the record — it queues the source for a human to review.
1,000
sources checked in the latest sweep
756
verified live
76% of all checked
0
confirmed broken — in the review queue
238
unverified (unreachable at last check)
often a bot/geo block, not a broken source
344
pages where we re-confirmed the record's own identifier is present
Sources last checked Sep 12, 2026 · 36 verdicts written in the last run. The authoritative full sweep runs from a US-based runner; results refresh here hourly.
✓ No source was confirmed broken at the last check. 238 sources could not be reached from the last runner and are being re-checked from a US-based runner.
Change ledger — what we observed, and when
Every law record and headline we track is stored in an append-only history: when a value changes, we write a new timestamped snapshot instead of overwriting the old one. That gives an auditable trail of what changed and when we first recorded it — shown here next to the source's own date, so you can see the observation lag. We have recorded 160,013 change-observations since Jul 7, 2026.
We would rather you catch a mistake than trust us blindly. If you report a verified factual error in our data — a wrong bill number, a mis-stated status or deadline, a broken or mislabeled source — and we confirm it, we'll credit you a free month of any paid plan (or extend an existing one) and fix it. Include the page URL, what's wrong, and the primary source that proves it.
One credit per account per quarter, and the credit goes to the first person to report a given error — issues already listed on this page don't qualify.
Primary sources. Records are built from primary government sources — statutes, bill text, official regulators — and each links back to its source. See our full methodology.
Source-liveness. A scheduled job re-fetches every cited URL from a US-based runner and classifies it: live, dead (404/410), a soft-404 shell, or unreachable/unverified. Only unambiguous failures are flagged.
Cross-checking. Where a source page renders server-side, we re-confirm that the record's own identifier is still present on it. Absence is never treated as proof of drift (many government sites render client-side), so it's queued for a human, not auto-changed.
No automated edits. A flag never rewrites a legal record. There is no LLM in this loop — every correction is made by a person against the primary source.
Append-only provenance. Every change is timestamped and kept; nothing is silently overwritten.
Not legal advice. AI Law Tracker is an information service, not a law firm, and its founder is not a lawyer. The data here is provided for general awareness and can contain errors or omissions; always confirm against the linked primary source and consult qualified counsel before acting. This page describes our verification process — it is not a warranty of completeness or correctness.