How GovPing works.

GovPing monitors 2,685 government and regulatory sources. Federal agencies, state regulators, courts, industry bodies, and selected international equivalents. Every source is added deliberately. We pick the canonical URL from the agency itself, not a mirror, aggregator, or secondary publication.

When an agency publishes to multiple surfaces (press release, rule docket, guidance portal), we track the surface most likely to contain the first detectable change. The source URL for every item links to that exact page.

Each source is checked on a schedule. The cadence depends on how often the agency actually changes the page. Daily for most press release feeds and dockets. More often for high-velocity sources like the Federal Register. Less often for slower-moving pages.

Detection is content-based, not timestamp-based. We compare the current page against the previous version and look for material content changes. A reformatted sidebar or a rotated testimonial is not a regulatory change, and our pipeline filters that out before anything reaches the index.

When the source is a PDF, we extract text, normalize it, and diff against the last known version. PDFs with scanned images go through OCR before comparison.

A detected change goes through an AI classification pass. The model reads the full content and returns a structured record in ORCA format: instrument type, regulatory stage, affected parties, jurisdiction, effective date, required actions, and a neutral summary.

We use large language models for this step. The model is prompted with the source page, the previous version where one exists, and a strict schema. The classifier's output is validated against the schema before anything is published.

Classification is explicitly not opinion. We describe what changed, not whether the change is good or bad. When an agency publishes a rule, we say the rule was published. We do not take positions on the rule.

Every change gets two AI-generated summaries. A neutral change summary describing what moved, and a "what changed" abstract describing the diff at paragraph level where we have both versions to compare.

Summaries are grounded in the source. The model is not permitted to add context that isn't present in the page. When a summary references a date, dollar figure, or named party, that data must appear in the source.

Summaries are not the source of truth. The source URL is. We publish the summary as a navigational aid and cite the original so readers who need the exact wording can go directly to the agency.

Every published change carries an ORCA record, a source URL, a detection timestamp, the effective date where available, and the issuing authority. Schema.org NewsArticle metadata lists the agency as the author, GovPing as the publisher, and the source URL as isBasedOn. This is the provenance chain journalists, researchers, and AI systems rely on to cite correctly.

RSS and OPML feeds are available for every source, category, jurisdiction, and role. Feeds are free and do not require an account.

We don't editorialize. We don't rank regulations by how important we think they are. We don't predict what the agency will do next.

We don't claim to cover everything. Some agency pages are impossible to monitor because the URL changes every week. Some PDFs are malformed badly enough that OCR can't recover them. Some sources are deliberately excluded because they publish aggregated content rather than original regulation. Our coverage reflects those choices and is visible by browsing the source list.

We don't replace primary sources. If an outcome depends on the exact wording of a rule, read the rule. GovPing is for discovery and triage, not for the legal record.

When we get something wrong, fix it. Misclassified instrument type, wrong jurisdiction, broken summary, missing effective date. Email [email protected] with the URL and what should change. We update the record and note the correction date.

For broader feedback on coverage or methodology, read our editorial standards or email [email protected].