How we verify

A DUI figure is publishable here only if it traces to a primary source: the state legislature’s own text of the statute, a court rule, or the licensing agency’s own publication. We do not source from other law-firm websites, aggregator pages, or summaries — including good ones. Those are where stale figures propagate, and copying one means inheriting its errors without inheriting its ability to notice them.

Each source is stored with its URL, its publisher, an authority tier, and a hash of the text as fetched. The hash matters: when a legislature amends a section, the hash changes, and the row goes back into review rather than sitting on the site looking current.

Extraction from a statute is automated. Publication is not. A proposed row — a penalty range, a hearing deadline, a licence consequence — sits in a review queue until an attorney accepts it. The acceptance writes that reviewer’s name and the date onto the row, and only rows carrying that stamp are served to the site, to the public data endpoints, or to assistants.

What the reviewer is confirming is narrow and checkable: that the statute says what we say it says, that the citation points where we claim, and that the version we read is the one currently in force. That is a documentary question, not an opinion about how any state’s law applies to anyone — which is why the reviewer does not need to be admitted in the state a given row describes. Every row also carries the source link, so a reader never has to take the review on trust.

The practical consequence is that this reference will always cover fewer jurisdictions than a site willing to publish unverified summaries. That is the intended trade. A missing state costs a reader one search; a wrong jail number costs them their planning, and costs us the only thing that makes the project worth doing.

When two of our own pages answer the same question differently, both are wrong until someone rules. Our canonical figures live in one file, and any value the network disputes is marked as unsettled and withheld from publication — it cannot reach a page or a structured-data block by accident, even though the competing versions are recorded so the disagreement stays visible.

This is not hypothetical housekeeping. A sweep of our own sites in August 2026 found the same Arizona jail question answered three different ways across domains we control. Every one of those answers came from a page that looked authoritative.

• We do not estimate. A value we have not verified is shown as absent, not as approximate.
• We do not average conflicting sources into a middle figure.
• We do not publish a jurisdiction because a competitor covers it.
• We do not present general information as advice about a particular case. The statutes are the same for everyone; the answer rarely is.

Roughly a quarter of official US legal-source hosts refuse automated retrieval outright — 21 of the 51 jurisdictions we map have at least one primary source that will not answer a machine, and several never answer at all. Those jurisdictions require a person to retrieve the document. It is the single biggest constraint on how fast coverage grows, and it is why coverage grows state by state rather than all at once.

Statutes also change on legislative schedules we do not control. A verification date tells you when we last looked; it does not promise nothing has changed since. For anything with a deadline attached, check the primary source we link — that is why we link it.

Tell us and we will fix it in public. See corrections for how we record changes, and editorial policy for who decides what gets published.