Journal · Our own computation

How much GLP-1 telehealth pricing is actually verifiable

An audit of the 82 pricing rows behind this site, classified by one deterministic evidence rule. The distribution below is computed at build time from the same manifest that renders our price tables — and it is not a flattering picture, including of us.

The distribution

What the distribution means

Each status answers one question: who is vouching for this number? "Provider reported" means the provider's own page said it — a real citation, but a mutable, promotional one. "Third-party reported" means we hold a source we cannot attribute to the provider's own domain; part of that class is an artifact of our own data hygiene, since only 1of 20 providers currently have an official_domain recorded, and without that field a provider's own page cannot be deterministically attributed to it. "Unverified" means no source URL at all — a number floating free of any citation. None of these classes tells you a price is wrong. All of them tell you nobody independent has reproduced it.

The inference we draw — labeled as inference — is that GLP-1 telehealth pricing operates on trust it hasn't earned mechanically: consumers compare figures that differ in evidentiary weight as though they differed only in dollars. That is the exact confusion our per-row badges exist to prevent, and it is why our rubric refuses "best price" designations built on unverified rows.

The structure of the prices themselves

Evidence status asks who vouches for a number; price type asks what kind of number it is. Our taxonomy classifies each row by structure, and the dataset breaks down like this:

Two readings. First, the commitment-tier share matters more than its size suggests: those rows are the prepaid bundles whose true cost depends on refund terms most providers don't disclose — the prepaid calculator quantifies what stopping early does to them. Second, the unconfirmed rows are figures we hold but could not classify structurally; they are excluded from every "cheapest" computation on this site rather than compared as though their structure were known. A market that published prices plainly would make both categories rare — their persistence is the subject of the companion piece on the entry-price illusion.

What this says about us

This audit indicts its own publisher. Our verification standard exists, is published, and is currently met by nothing — because meeting it requires human labor (open the source, reproduce the claim, archive it with a checksum, sign it) that has not yet been done for a single row. We could have shipped inherited "verified" labels; the dataset arrived with some, and wedemoted them all rather than wear trust we didn't earn. Until rows start clearing that bar, this site's honest value is the labeling, not the verification.

What would change this picture

The distribution is not fixed; specific actions move specific rows. Recording each provider'sofficial_domain — pure data entry — would let own-domain sources reclassify from third-party to provider-reported, resolving the attribution artifact noted above. Capturing a source URL for each unsourced row would eliminate the "unverified" class row by row. And the top of the ladder has exactly one entrance: a human opens the source, reproduces the claim, archives a checksummed capture, and signs the review record — at which point, and at no point earlier, a row earns "verified by us." We publish this list because it doubles as our own backlog, in priority order, and because a reader can hold us to it: if a future build's distribution hasn't moved, we didn't do the work.

Method

Every figure on this page is computed at build time from provider-pricing.json — the same manifest that renders our medication price tables — via the same normalizeStatus() rule every page uses. No number here is hand-typed, so this analysis cannot drift from the tables it describes; if the manifest changes, the analysis recomputes on the next build. The classification rule itself is documented in theevidence policy.