We would rather be argued with on a specific line than trusted in general. This page sets out every line, in the order we work through it, so that a disagreement can point at the step where it arises.
Part one: eight checks, each stamped
Every file opens with the same ledger. Each check is answered from the storefront's own pages or from a public record, and the stamp is one of three words: Verified, Partial, Not verified. A fact we could not confirm is stamped Not verified; we do not round silence up.
- Third-party laboratory testing. Verified if the supplier states that an outside laboratory tests its products; Partial if the claim covers only part of the catalog; Not verified if testing is denied, absent or unclear.
- Certificates of analysis open to the public. Verified only if we opened a certificate with no account and no order.
- Certificates tied to a specific lot. Verified if the certificate carries a lot identifier that matches a listing.
- Testing laboratory named. Verified if the certificate or the testing page names the laboratory that did the work.
- Research-use-only disclaimer present. Verified if the storefront labels its material for laboratory research use only.
- No FDA warning letter on record. Not verified if the US Food and Drug Administration has published a warning letter to the company.
- Identifiable company. Verified with a street address or a phone number; Partial with an email address only; Not verified with a web form only.
- Currently operating. Not verified if the storefront has closed.
The tally, written as "6 of 8", is printed on every file and in the registry. It is a plain count; it is not weighted and it is not the seal.
Part two: four weighted lines
The ledger says what is missing. The seal says how much it matters. Six lines are each scored from 0 to 100 out of the record; four of them carry weight on this desk and two carry none.
- 45percentTesting and certificatesOutside testing stated (yes 40 points, limited 18, no claim 6, denied 0); a certificate that opens for anyone, 15; a certificate tied to a lot, 12; one laboratory named, 8, or two or more, 14; a laboratory that appears to hold ISO/IEC 17025 accreditation, 6; methods printed on the certificate: HPLC 3, mass spectrometry 4, endotoxin 4, sterility 4. Capped at 100.
- 35percentTransparencyResearch-use labeling, 22; a published email, 14; a phone number, 12; a street address, 18; a stated location, 10; a certificate page that resolves, 14; a stated founding year, 6; a written returns policy, 4.
- 15percentTrack recordYears trading as stated (under two, or unstated, 12; two or more, 20; four or more, 30; seven or more, 40); a trust-site figure scaled by its volume, up to 35, or 8 where none exists; up to 15 for facts in the company's favour and up to 24 off for serious facts against it; 35 off for a public FDA warning letter; 5 or 10 for a forum footprint. A closed storefront scores zero here.
- 5percentOrdering termsBase 10; card acceptance 25; each further payment method 5, up to 20; a stated dispatch time 15; shipping abroad 10; named carriers 8; a returns policy 7; a guarantee 5.
- 0percentCatalogRecorded in the source record, scored, and multiplied by zero. Breadth says nothing about truth.
- 0percentPriceRecorded in the source record, scored, and multiplied by zero. A discount does not make a certificate more real.
The weighted sum, rounded, is a number from 0 to 100 that we keep in the dataset and do not print on the file. The letter is read off it.
Part three: two rules that override the sum
- A public FDA warning letter holds the number at 59 or below, which is a C. However complete the rest of the ledger, a regulator's letter on file keeps a storefront out of the B and A bands until the letter leaves the public record.
- A storefront we could not reach gets no seal. If the site did not load, redirected elsewhere or published nothing we could open, the number is set to zero and the file reads "no seal". A closed storefront is likewise filed without a seal.
Part four: the bands
| Number | Seal | What the desk means by it |
|---|---|---|
| 90 to 100 | A+ | Every material claim opened to a document. |
| 83 to 89 | A | Claims hold; small gaps in the record. |
| 76 to 82 | B+ | The testing claim verified; one notable gap elsewhere. |
| 68 to 75 | B | Core claim verified; the company is thin on identity or history. |
| 60 to 67 | C+ | Mixed. Open the certificate for the lot in hand before relying on anything. |
| 50 to 59 | C | Material claims unverified, or an FDA letter holds the file here. |
| 40 to 49 | D | Few claims could be opened. |
| Below 40 | F | The record does not support what the storefront says. |
What we leave out on purpose
- Our own laboratory results. We run none. A document you can open yourself is worth more than one you would have to take on our word.
- Reviews, testimonials, stars. None collected, none printed, none marked up. See the editorial policy.
- Any statement about use. Every storefront in the registry sells material labeled for laboratory research only. No check, line or sentence here concerns any other use.
- Referral status. Whether a link out earns this site money is disclosed and is invisible to the formula.
- Rankings. We do not sort files against one another. The registry is alphabetical and the side-by-side pages declare no winner.
Where the fields come from
Each file is assembled from the storefront's public pages (home, product, testing, about, contact, shipping) and from public records (the fda.gov warning-letter database, the trust-site profile the storefront links to, business registries). Each file lists the pages read. Fields are re-read on a rolling basis; the whole registry regenerates when a field changes and every page shows the month. The dataset is on the data page as CSV and JSON so that anyone can press every seal again without us.
Reader questions
Answered from the fileCan a supplier pay for a better seal?
No. The same fields go through the same formula for every file. The only lever a supplier holds is to change a fact we can open, for instance by publishing lot-tied certificates where anyone can read them.
Why do you give catalog size and price no weight?
Because neither tells us whether a claim is true. A large catalog is easy to assemble and a low price is easy to print. We still record both in the source record so a reader can see them; they simply do not move the seal.
Why not use customer reviews?
In this trade review counts are bought, lost and deleted, and several storefronts have had trust-site profiles removed outright. Where a profile exists we record its figure and the date we read it as a fact in the track-record line. We never collect reviews, never quote them and attach no rating markup to any page.
Do you test anything yourselves?
No. We check what a supplier publishes against what regulators and registries publish. A certificate from a named outside laboratory for a specific lot is the strongest evidence we accept, and producing it is the supplier's job.
How is a fact corrected?
Send editors@peptidetruth.org the address of the page that shows the correct value. The field changes at the next regeneration and the checked date on the file moves.