The short version: a certificate of analysis [COA] worth trusting covers five measures — net content [measured milligrams], identity [LC-MS, liquid chromatography–mass spectrometry], purity [HPLC-UV, high-performance liquid chromatography with ultraviolet detection], endotoxin [USP <85> — United States Pharmacopeia chapter 85; angle brackets denote a numbered pharmacopeial chapter], and a microbial screen, read in that order — names the lab that ran them, and carries an accession number [the testing lab's own identifier for one specific analysis] you can verify with that lab.
Most certificates on the market cover one to three of the five. "Purity 99%" alone is not a quality program. It is one number.
- Purity is a peak-area ratio — it says one substance dominates the vial. On its own it does not confirm which molecule, how many milligrams, or whether it is clean.
- Net content is the number that costs you money. A vial labeled 50mg that measures 45mg is a 10% price increase nobody announced. Good labs print the measured fill.
- Endotoxin and sterility are different tests. USP <85> [endotoxin] and a PCR microbial screen are common. USP <71> [true sterility] almost never is — treat any casual "sterile-tested" claim with suspicion.
- A certificate with no named lab and no accession number cannot be verified and should be treated as decoration.
- The single best question to ask any vendor: "was this test run on your lot, in your name?" Manufacturer batch data recycled from a supplier is not the same document.
Why this beats reviews
Vendor reputations change faster than you can check them. A certificate can be read in five minutes and cannot change its story afterwards.
Vendor reputations move fast in this market — companies appear, take orders, and disappear inside a year, and review threads are gamed in both directions.
Certificates don't argue. If you can read one, you can evaluate any vendor in the time it takes to open a PDF, regardless of what Reddit thinks this month.
The problem is that "third-party tested" has been flattened into a slogan. It can mean a full five-measure panel run on the vendor's own lot at a named US lab.
Or it can mean a purity screenshot recycled from an overseas supplier's batch, tested on someone else's vial, months before your order existed. Both get the same badge on a product page. The certificate is where the difference lives.
The five measures
Net content, identity, purity, endotoxin, microbial — read in that order, because each one catches a failure the others cannot see.
| measure | method | what it tells you | what it can't tell you |
|---|---|---|---|
| Net content | gravimetric / assay | The measured milligrams actually in the vial, against the label claim. | Anything about what the milligrams are. |
| Identity | LC-MS | Mass confirmation that the molecule is the stated compound. | How much of it, or how clean. |
| Purity | HPLC-UV | What fraction of UV-absorbing material is a single substance. | Which substance, the fill, contamination. |
| Endotoxin | LAL [limulus amebocyte lysate] · USP <85> | Bacterial endotoxin [pyrogen] load against a threshold. | Sterility. Different test entirely. |
| Microbial | PCR screen | Whether microbial DNA is detectable. | Sterility — DNA-absence is not a culture result. |
1. Net content, unarguable
Start here, not at purity. Net content is the one measure a purity percentage absolutely cannot catch, and it is the one with a direct price on it.
A 50mg label over a 45mg fill is a 10% surcharge. A 50mg label over a 40mg fill is 25%. Underfill is common enough across the market that independent testing groups treat fill accuracy as a standard pass/fail criterion alongside purity.
What good looks like: the certificate prints a measured number — "net content: 48.22 mg" — rather than a checkbox.
A vendor willing to publish a measured fill, including when it lands slightly under label, is showing you real data. A vendor who only ever shows round numbers that exactly match every label is showing you a template.
2. Identity, not purity
Here is the nuance most buyers miss in both directions. HPLC purity runs are done against a reference standard, so a peak at the expected retention time is presumptive identity evidence. Purity is not identity-blind.
But retention time cannot separate close analogs: a truncated sequence or a similar peptide can co-elute at the same time and still read "99%." That is exactly why LC-MS — mass confirmation — is a separate, separately-paid line on serious certificates.
If the purity number settled identity, nobody would pay for the mass spec.
3. Purity, insufficient alone
Purity by HPLC-UV is a peak-area ratio: of the material the detector can see, what fraction sits in one peak. It is a real and important measure — a 92% purity peptide carries 8% of something else.
But it is the cheapest number to produce and the easiest to fake, which is why it is the one number every vendor shows. When a certificate is purity-only, the honest reading is: this vendor bought the minimum.
4. Endotoxin, the injectable question
Bacterial endotoxins are pyrogens — fever-inducing fragments of gram-negative bacterial cell walls. The standard test is the LAL assay per USP <85>, reported against a threshold [good labs run it in duplicate].
For anything in injectable format, this is not an exotic extra — it is the difference between a chemistry document and a safety document. Very few research vendors run it. The ones that do tend to say so loudly, because it is genuinely rare.
5. Microbial, and sterility
This is where the market's language gets slippery, so read carefully:
- USP <85> is an endotoxin test. It detects pyrogens.
- A PCR microbial screen looks for microbial DNA. "No detectable microbial DNA" is a contamination screen — useful, real, limited.
- USP <71> is the actual sterility test — a roughly 14-day culture-based method that consumes sample units. It is expensive, slow, and almost never run on research peptides.
DNA-absence is not sterility: PCR can miss viable organisms below its detection floor and can equally flag dead DNA. So when a vendor's marketing says "sterility tested," open the certificate and look for USP <71>.
In our experience reviewing certificates across the market, it is almost never there — the claim is usually endotoxin or PCR wearing a bigger word.
Note the legitimate exception: "sterile" as a form descriptor ["sterile diluent · USP" on bacteriostatic water] describes how a product is filled, not a test result on a peptide lot.
The step nobody takes
A real certificate names the lab, prints an accession number and matches your lot. Those three fields let you confirm it without asking the vendor.
A real third-party certificate carries three pieces of metadata that make it checkable:
- The lab's name — a real, findable laboratory, not "our testing partner."
- An accession or reference number — the lab's own identifier for that specific analysis.
- The lot number — matching the lot printed on the vial you received.
Reputable peptide-testing labs offer verification — some print lookup codes directly on the certificate. Take sixty seconds and check. A certificate that cannot be verified with the lab that supposedly issued it is, functionally, artwork.
And a striking pattern we keep finding: vendors whose homepage claims a long list of tests. Their own lab-results page shows fewer, or shows none at all, just prose about testing. When the claim count and the document count disagree, believe the documents.
Whose vial was actually tested
Check the client field. A certificate in the vendor’s own name covers the vial you bought; supplier batch data covers powder from months earlier.
The quietest distinction in the whole market: tested in whose name, on whose lot. Bulk peptide arrives from a synthesizer with the manufacturer's batch data.
A vendor can forward that paperwork untouched — which tells you about a batch that existed upstream, months ago, before dilution, handling, and refilling. Or the vendor can pull vials from their own finished lot and pay a lab to test those, with the vendor's name as the client on the certificate.
The second document is categorically stronger. It covers the product as sold, not the powder as bought. When you see a certificate, check the "client" field — if it names some other company, you are reading recycled supplier data.
Ask the vendor directly: was this run on your lot, in your name? The speed and specificity of the answer is itself a signal.
Six disqualifying red flags
Six certificate patterns are disqualifying on sight, and every one of them is visible before you order.
- No certificate at all — or "COA available on request" that never materializes.
- No named lab, no accession number — unverifiable by construction.
- "GMP-standard facility" — weasel phrasing chosen because "GMP-certified" would be falsifiable. Real certification comes with a cert number and an issuing body.
- Sterility claims without USP <71> on the document.
- Every lot, every compound, always 99.9% — real analytical data has texture. Perfect uniformity is a template tell.
- A homepage test list longer than the certificate's — count both. They should match.
What good actually looks like
A complete certificate names the lab, prints the accession number and lot, reports all five measures, and shows the measured fill even when it lands under label.
A handful of vendors now publish the full document: named lab, accession number, lot, the five measures. The measured fill is printed even when it lands a few percent under label.
That last detail is the strongest trust signal in the market. Publishing a 48.22 mg measurement against a 50 mg label is something a fabricator would never do.
Our 2026 COA audit grades vendors on exactly this axis, with links to the documents themselves. The criteria — and our referral disclosure — are in the editorial policy.
FAQ
Is a COA legally required for research peptides?
No. Research-use-only compounds sit outside the drug approval framework, and no regulation compels a vendor to publish certificates. That is exactly why the ones who voluntarily publish full panels are worth noticing — the market's floor is zero.
Nothing forces a vendor to measure net content, run an LC-MS identity check, pay for a USP <85> endotoxin assay, or name the lab that did the work.
Every certificate you see is a choice somebody made and paid for. So the absence of a certificate proves nothing legally and a great deal commercially. A vendor who will not publish has decided the document is not worth the cost of producing it.
Judge the market on what it volunteers. The vendors who print a measured fill, a named lab, an accession number and a matching lot number are operating far above a floor that starts at nothing.
Can I trust a COA hosted on the vendor's own site?
Hosting location matters less than verifiability. A certificate on the vendor's site that names the lab and carries an accession number you can check with that lab is strong. A certificate hosted anywhere that cannot be verified is weak.
Three fields decide it. The lab's name, which has to be a real findable laboratory rather than a vague reference to a testing partner. An accession or reference number, the lab's own identifier for that specific analysis.
And a lot number matching the lot printed on the vial you received. Reputable peptide-testing labs offer verification, and some print lookup codes directly on the certificate, so the check runs in about sixty seconds.
While you are on the site, count the tests the homepage claims and count the tests the certificate documents. When the two counts disagree, believe the documents.
What purity number should I look for?
Serious vendors and independent testing groups generally treat ≥99% by HPLC as the standard for common research peptides. But read the rest of this article — a 99% purity number with no identity, fill, or endotoxin data is a one-fifth answer.
Purity by HPLC-UV is a peak-area ratio: of the material the detector can see, what fraction sits in one peak. A 92% peptide carries 8% of something else, so the number is real and worth having.
It is also the cheapest number to produce and the easiest to fake, which is why it is the one number every vendor shows. It cannot tell you which molecule is in the vial, how many milligrams arrived, or whether the contents are clean.
Watch for the opposite failure too. Every lot, every compound, always 99.9% is a template tell — real analytical data has texture.
What is an accession number?
The testing lab's internal identifier for one specific analysis of one specific sample. It is what lets you go around the vendor and confirm the document with the lab directly.
No accession number, no verification path. It is one of three pieces of metadata that make a certificate checkable. Those are the lab's name, the accession or reference number, and the lot number matching the vial you received.
All three have to be present, and the lot has to match. Reputable peptide-testing labs offer verification, and some print lookup codes on the certificate itself, so the check takes about sixty seconds.
A certificate that cannot be verified with the lab that supposedly issued it is, functionally, artwork. No named lab and no accession number is one of the six certificate patterns that end the conversation before you order.
What should a good peptide COA include?
A strong certificate of analysis covers five measures, and the order they are read in matters. They are net content as measured milligrams, identity confirmed by LC-MS, purity by HPLC-UV, bacterial endotoxin by the LAL assay per USP <85>, and a microbial screen.
Net content comes first because it is the one measure with a price attached and the one a vendor cannot argue with. It also names the testing lab, carries a lot number and accession number, and can be verified with the lab directly.
Identity sits apart from purity because retention time cannot separate close analogs, and only mass confirmation settles which molecule is in the vial. Endotoxin sits apart from sterility because USP <85> detects pyrogens while USP <71> cultures for living organisms.
The strongest documents print the measured fill even when it lands under label: 48.22 mg against a 50 mg claim is a number a fabricator would never publish.
Does 99% purity mean the peptide is what the label says?
Not by itself. HPLC purity is a peak-area ratio. It says one substance dominates the vial, but a percentage on its own does not confirm the molecule. That is what the separate LC-MS identity test is for.
The nuance runs in both directions. HPLC purity runs are done against a reference standard, so a peak at the expected retention time is presumptive identity evidence. Purity is not identity-blind.
But retention time cannot separate close analogs. A truncated sequence or a similar peptide can co-elute at the same time and still read 99%. LC-MS confirms mass, which is why it appears as a separate, separately-paid line on serious certificates.
If the purity number settled identity, nobody would pay for the mass spec. A purity figure also says nothing about how many milligrams are actually in the vial.
Does a COA prove a peptide is sterile?
Almost never. True sterility testing is USP <71>, a roughly 14-day culture-based test that is rarely run on research peptides. Endotoxin testing [USP <85>] and PCR microbial screens are different tests and do not establish sterility. USP <85> detects pyrogens — fever-inducing fragments of gram-negative bacterial cell walls.
A PCR screen looks for microbial DNA, and DNA-absence is not sterility: PCR can miss viable organisms below its detection floor and can equally flag dead DNA. USP <71> is expensive, slow, and consumes sample units, which is why the research market skips it.
So when marketing says sterility tested, open the certificate and look for USP <71>. One legitimate exception: sterile used as a form descriptor on bacteriostatic water describes how a product is filled, not a test result on a peptide lot.
How do I verify a COA is real?
Take the accession or reference number printed on the certificate and check it with the testing lab directly. Reputable third-party labs offer lookup or verification. A certificate with no named lab and no accession number cannot be verified and should be treated as artwork.
Three fields make a certificate checkable. The lab's name, which has to be a real findable laboratory rather than a vague reference to a testing partner. The accession or reference number, the lab's own identifier for that specific analysis. And the lot number, matching the lot printed on the vial you received.
Some labs print lookup codes directly on the document, so the check takes about sixty seconds. Then read the client field. If it names another company, you are looking at recycled supplier batch data rather than a test on the vendor's own finished lot.
References
- USP General Chapter <85> Bacterial Endotoxins Test — usp.org
- USP General Chapter <71> Sterility Tests — usp.org
- FDA, Facts About the Current Good Manufacturing Practices [CGMP] — fda.gov
- FDA Import Alert 66-80, Detention Without Physical Examination of GLP-1 Bulk Drug Substances — accessdata.fda.gov · covered in depth in our supply-chain report
disclosure: inside your peptides is published by the owners of the next lab, the only vendor we are paid by, including the next lab. grades and rankings follow the published criteria in our editorial policy — never referral terms. research + education only · not medical advice.