Every alert published by StateLicensure has been reviewed by a human editor. No diff-to-publish automation, no AI-generated summaries pushed to production. Here's exactly why and how that works.
The problem with auto-publishing
State medical board websites are not built for machine consumption. Rule changes appear in PDFs, in footnotes, in administrative notices that reference other notices. A rule change detected by a crawler is often partial — it captures that something changed, not what it means, whether it's substantive, or what action (if any) a physician should take.
Auto-publishing a detected change as an alert introduces noise. A physician who receives five low-signal alerts starts ignoring all of them, including the ones that matter. We'd rather send fewer alerts that are always worth reading than many that train you to ignore us.
We'd rather send fewer alerts that are always worth reading than many that train you to ignore us.
What our editors actually do
When our monitoring detects a change on a source site, the change enters a review queue. An editor reads the source material, evaluates whether the change is substantive, writes a plain-English summary that explains what changed and what — if anything — a physician needs to do, and attaches the primary-source citation.
The editor then assigns a severity tier: Informational (worth knowing, no action needed), Action needed (review required before your next relevant encounter), or Urgent (compliance deadline within 7 days or immediate practice impact). Only then does the alert publish.
What we use AI for
We use AI to help detect changes across sources — crawl scheduling, diff detection, and flagging potential changes for review. AI does not publish alerts autonomously. The editorial judgment is human.