Methodology

Source tiers

Not every source carries the same weight. We rank what we read, and a status is only ever as strong as the best source under it.

Primary sources speak.

A status rests on primary material: the legislative or statutory text, an official government or regulator publication, a ballot measure as certified, or a court record. When a record says psilocybin is decriminalized in a place, the citation goes to the instrument that made it so.

Credible reporting corroborates.

Established news outlets, legal analyses, and academic writing are used to corroborate and to date a change, never as the sole basis for a status. Where a secondary source is the only thing available, the entry stays in verification and says so.

Aggregators only help us find.

Bill-tracking services and legal-status aggregators point us toward what changed and when. We then read the primary instrument and cite that. Aggregated status tables are a finding aid, not a source of record, and they are never republished here as the answer.

Verification cadence

Every entry goes through a source-verification pass before it is published, and each record shows the date it was last reviewed. After that, re-review runs on two cycles, because the law does not move at the same speed everywhere. The 10 jurisdictions we watch most closely are re-verified on a 60 day cycle: Colorado, Illinois, Massachusetts, Minnesota, New Mexico, Oregon, Rhode Island, South Dakota, Virginia, and Washington. Every other record in the register is re-verified on a 180day cycle. On top of both, a record is pulled forward as soon as our monitoring picks up movement in it, so a bill or a ruling can bring a review months early. When a status changes, the change is logged, dated, and typed in the record’s own changelog.

A cycle says how often we come back to a record. It is not a claim that any particular entry was checked today. The last reviewed date on each record is the authority: it is when that entry was actually read against its sources, and it is the date to cite. The watch list above changes as legislatures do.

Library and research corpora are stamped with the date they were last swept, currently 14 Jun 2026, and each library entry carries its own record-reviewed stamp so freshness is visible per item rather than claimed for the collection as a whole. Link checking is separate and narrower: one automated sweeps have run, each library and research page states the date of the check it carries, some publishers refuse the sweep, and no entry on this site claims a hand check.

AI and machine assistance

The sources above are watched by software we wrote. It runs over the legislature pages, gazettes, court dockets and trial registries in scope, compares what it finds against what it found last time, and flags anything that moved. A flag is a lead. Nothing is published on the strength of one.

Language models do some of the reading that follows. They work through long instruments, pull out the dates and section numbers, and draft the summaries and explanatory copy a reviewer starts from. That scope is stated broadly on purpose. A model has touched the first version of much of the explanatory prose on this site.

Models fail in a particular way. They produce a sentence that reads correctly, carries a plausible citation, and has no document under it. So every status, date and citation is checked against the primary source the record cites, and a claim that cannot be traced back to a document does not go on the site.

No new entry, correction, or status change goes live without a person approving it. That is the claim we can prove: approval is a step in the publishing path, and nothing reaches the site around it. We do not claim that a human eye has read every line of every source. We could not show it.

Downstream of that approval the path is mechanical. The change is written to the register, and the figures this site prints about it are counted at build time, so a number on a page and the data behind it cannot drift apart.

How AI systems and other automated readers may reuse what comes out is a separate question, and the license page answers it.

Confidence rubric

Each record carries one of four verification marks, printed beside the date it was last reviewed. The mark is a claim about our evidence, not about the law, and we would rather show a weaker mark than overstate what we can prove.

Checked against the cited sources.

At the review date shown beside it, the status was read against the primary or corroborating material the record lists, and that material still supports it. This is the mark most records carry.

Corrected and re-checked.

An earlier version of the record was wrong. The correction is a dated, typed line in the changelog on that record, and the status was re-read against its sources at the same review.

Status changed.

The law itself moved, rather than our reading of it. The change is logged and typed in the changelog, carrying the status it moved from and the status it moved to.

Reviewed, with source verification pending.

The entry was read at that review, but a primary source is still missing or the sources on file do not yet fully agree. The record carries the weaker mark openly rather than overstating what we can prove.

Corrections policy

We correct in the open. There are no silent edits: when a status, date, or citation is wrong, the fix is entered as a visible, typed changelog line inside the affected record, so the history of what we said and when we corrected it travels with the record itself. If you find an error, the way to report it and what happens next are set out on the terms page.

Legislative data credit

Some of the pending-legislation and bill-tracking data used to find and date changes is drawn from LegiScan, made available under the Creative Commons Attribution 4.0 license (CC BY 4.0). We use it as a finding aid, in keeping with the rule above: LegiScan and services like it tell us a bill moved, and we then read and cite the primary legislative text. We credit LegiScan wherever its data informs a record, and no provider’s raw feed is republished here. That boundary is written into what we license: the dataset licensed here comprises facts independently verified against primary sources; it does not include or sublicense LegiScan-provided content. The grant itself is on the license page.

Citation guide

You are welcome to cite the record. Every record page carries a “Cite this record” block with a ready-to-copy citation, but if you are writing one by hand, the elements are the jurisdiction and what the record is about, the publisher, the accession identifier where one is shown, the as-of date the status was current, and the record URL.

Format. Jurisdiction: psilocybin legal status. The Psilocybin Record. Accession ID. Status as of [house date]. Record URL.

Example. Oregon: psilocybin legal status. The Psilocybin Record. Accession US-OR-001. Status as of 17 Aug 2026. https://psilocybinlegalization.com/united-states/oregon

We ask that you cite the as-of date alongside the status, because the status is only true as of the date it was checked, and drug law moves.

Library inclusion criteria

The library collects third-party videos, books, and articles for context. It is a reading list, not a claim: nothing in the library is an endorsement, and no item makes or supports a medical or efficacy claim. Items are included on a few plain rules.

• Video is embedded only from the rightsholder’s own official channel. Subscription-only titles are linked, never rehosted.
• Each item is classified by authority, peer-reviewed, institutional, journalistic, or practitioner, with its publisher and year shown. Practitioner works are marked a practitioner reference, not medical evidence.
• We carry no affiliate links and take no commission on anything in the library. A rights holder can ask us to remove an item and we will.
• How-to material, dosing, sourcing, or step-by-step protocols, is out of scope and is not listed.

The video shelf shows this filter at work. A starting pool of 198 candidate videos was narrowed to 89: we removed unofficial re-uploads, duplicates, titles that were unavailable or could not be embedded from an official channel, low-authority clips, off-topic material, and anything that read as how-to. What remained is what a rightsholder actually publishes and what a reader can actually reach.

Study entries carry two things we do not write ourselves: a link to a free lawful copy where the publisher or an archive has posted one, and any correction, erratum, or retraction the publisher has since issued against the study. Both are read from the open bibliographic record, and a notice is shown whether or not it changes what the study concluded. Metadata via OpenAlex (CC0) and Crossref.

The register and any analyses never cite library items as evidence, and personal accounts, including trial testimony, are presented as individual experience, not proof. Evidence lives in the primary sources; the library is there to help you read around them.