Peptide Research Literature Index
A structured, dated index of the research-peptide landscape

Does a page have to promise something before FDA calls it a drug claim? On one day FDA wrote to fourteen websites selling the same drug, and the weakest evidence it quoted was two sentences framed entirely as research

Reference page · published 2026-10-20

Disclosure: this index is operated by Artemis Labs, which sells research compounds to laboratories. This page is about the kind of sentence we ourselves write — it is not a page about somebody else’s problem, and the section below on why this page exists at all says so with our own pages measured. Artemis Labs has no connection to any company named here, sells none of the products those companies sold, and sells no controlled substance of any kind. Nothing here is legal advice. Every quotation is verbatim from an FDA warning letter, an FDA record export, or the Code of Federal Regulations, each of which was fetched and read for this page on October 20, 2026 and each of which is listed under Sources. Three limits belong at the top. The companies described here received warning letters, which are agency allegations opening a correspondence, not findings of a court, and none of the fourteen has a close-out on the record. The compound in these letters has an approved human counterpart and is a scheduled controlled substance, which — as the counter-evidence section sets out at length — is a real limit on how far any of this carries to compounds that have neither. And the sales copy quoted here is reproduced because it is the matter FDA charged, not because any of it is true.

Short answer: no. A promise is one way to show intent, not the test for it. The regulation FDA cites in all fourteen of these letters, 21 CFR 201.128, says intended use may be shown by a seller’s expressions, the design or composition of the article, or by the circumstances surrounding the distribution of the article. Nothing in that sentence requires a claim. And the letters bear it out: alongside pages that promised relief from named conditions, FDA quoted a sentence giving a compound’s chemical name and what it does to perception; a sentence giving its drug class and the decade it was developed; a product title of three words; and, in the one letter signed by the office that writes to research-compound sellers, two bullets whose grammatical subject is researchers. That letter also records, in a footnote, that the site carried the words for research and scientific use and for non-clinical purposes, and says the labeling establishes human-drug intent despite them.

If you run a research-compound website, you have written three kinds of sentence that are not sales copy and never felt like claims. The one that says what the molecule is. The one that says what class of thing it belongs to and when it was first made. And the one that says what researchers use it to study. Those are the sentences a careful seller writes instead of claims — they are what you fall back on when you have decided not to promise anything.

This page asks a narrow question about them, and answers it from documents, not from reasoning: when FDA says a website is selling an unapproved human drug, what does it actually put in the letter as the evidence? We picked a set where the comparison is unusually clean. On a single day, FDA sent fourteen letters about the same compound, with the same subject line, charging very nearly the same sections. The product is held constant. What varies is the writing on the websites — and therefore what FDA chose to quote.

The set, and why it is a fair comparison

On June 23, 2026, the Center for Drug Evaluation and Research issued fourteen warning letters, every one of them addressed to a website selling ketamine. We did not take that count from anybody’s summary. We pulled FDA’s own export of its warning-letter list and counted the rows: fourteen dated 06/23/2026, all fourteen carrying the identical subject value Unapproved New Drugs/Misbranded. All fourteen were posted to fda.gov 14 days later, on July 7, 2026.

Two columns in that export are worth reading for what they do not contain. As of today — 119 days after issue — none of the fourteen rows carries a response letter, and none carries a close-out. That is a measurement and not an assumption: in the same agency export queried for a different term, 5 of 46 rows do carry a close-out date, so the column populates when there is something to populate it with.

The letters split by office, and the split is the reason this page exists:

Office within CDERLettersSigned by
Office of Drug Security, Integrity, and Response (Internet Pharmacy Task Force) 13Sangeeta Vaswani Chatterjee, PharmD, Director
Office of Unapproved Drugs and Labeling Compliance 1 — MARCS-CMS 728919 Tina Smith, MS, Captain, U.S. Public Health Service, Director

We read all fourteen signature blocks to get that table. The second row is the one that makes this a research-compound story and not a ketamine story. The Office of Unapproved Drugs and Labeling Compliance is the office that writes the letters to research-compound sellers, and Captain Smith is the officer who signs them. That is not inference either: of the pages in this index that record a warning letter to a research-compound seller, 13 name her as the signatory. One of the fourteen letters in this cluster came from her desk. It is the only one whose quoted evidence is research framing, and it is the only one of the fourteen that does not recite the compound’s controlled-substance schedule.

What FDA quoted from the research-framed page — all of it

Letter 728919 went to a site reached at buyketaminepowderonline.com. Here is the complete evidence paragraph. We are quoting the whole thing, lead-in included, because the length is the point:

Based on a review of your website, your ketamine powder is a drug under section 201(g)(1) of the FD&C Act (21 U.S.C. 321(g)(1)), because it is intended for use in the diagnosis, cure, mitigation, treatment, or prevention of disease, and/or intended to affect the structure or any function of the body. [a footnote marker, number 4, falls here in the letter; its text is quoted further down this page] Examples from your website labeling that provide evidence of the intended use (as defined in 21 Code of Federal Regulations (CFR) 201.128) of this product as a drug include, but may not be limited to, the following:

Neuroscience and Brain Function Research . . . Researchers use Ketamine to model dissociative states and altered neural connectivity, helping to explore cognition, perception, and consciousness at a molecular and systems level.

Psychiatric and Mood Disorder Research Models . . . Ketamine is frequently studied in preclinical and mechanistic research related to depression and mood regulation, particularly in examining rapid-onset neuroplastic changes.

That is the entire evidentiary showing in the letter. There is no third bullet. Read the two bullets as sentences and notice what they do not do. Neither promises an outcome to a reader. Neither addresses a reader at all: the grammatical subject of the first is Researchers and of the second is the compound, in the passive. The first names no condition whatsoever — its nouns are cognition, perception and consciousness. The second does name depression and mood regulation, but as the subject matter of preclinical and mechanistic research, which is a true description of where that literature sits.

We also measured what is absent from that letter, because a reader is entitled to check whether some other claim was doing the work and simply went unquoted. Inside the body of 728919 there is no price, no customer review, no testimonial, no reference to a social media account, and no purity or analytical claim. Across all fourteen letters in the cluster the same probes return 0 of 14 for testimonial, social media, purity, certificate of analysis, and for every payment term we tried; the live controls in the same pass return 14 of 14 for misbranded, unapproved and website, and a deliberately meaningless control returns 0. So the two bullets are not a summary of a larger case. They are the case.

The footnote that disposes of the disclaimer

The site was not silent about what it was for. Footnote 4 of the letter records what it said, and answers it in the same sentence:

Despite statements on your website labeling marketing your product “for research and scientific use” and “for non-clinical purposes,” evidence obtained from your website labeling establishes that your product is intended to be a drug for human use.

This index already has a page on that override sentence and how consistently FDA has used it — see what a research-use-only notice does not do. What this letter adds to that record is a third grammatical form of the disclaimer. The familiar forms either assert research use (for laboratory and research use only) or deny human consumption. The phrase for non-clinical purposes does neither: it defines the exemption by negating the clinical setting. It was overridden with the same sentence as the others.

What FDA quoted from the other thirteen, and why thirteen-claims-against-one-research-page is too tidy a story

The clean story would be that the thirteen Task Force letters quote disease claims and the fourteenth quotes research framing. We expected that, and our own measurement did not support it. It is closer to a gradient, and the interesting part is how far down the gradient goes before the charge stops changing — which it never does.

Our counting rule, stated so you can discount it: we extracted every quoted run of four words or more that appears after FDA’s own evidence lead-in (includes the claim, or include, but may not be limited to, the following) and within that paragraph. That yields 30 quoted strings across the fourteen letters, with every letter contributing at least one. The rule is crude in two directions and we would rather say so than round. It undercounts: letter 725156’s evidence includes the three-word phrase Mental Health Treatments, below our threshold. And a first pass overcounted, by sweeping up a sentence of FDA’s own prose as though it were site copy; the rule above exists because of that error, and it now runs with a control that fails if that sentence reappears.

Here is the gradient, in FDA’s own quotations. At the top, pages that promise:

Those need no analysis. They name conditions and promise results, and nobody writing a research page writes them. Further down, the sentences stop promising:

And at the bottom of the gradient are the three that matter to anyone writing about compounds:

A sentence that says only what the molecule is

From letter 725151. FDA’s complete quoted evidence for the first of two products on that site:

Ketamine ((RS)-2-(2-chlorophenyl)-2-(methylamino)cyclohexanone) is a powerful psychoactive drug with the unique ability to decouple one’s conscious experience from the body.

Read it as a grammatical object. It contains a systematic chemical name, a drug-class noun, and a description of an effect on a person. It names no disease, offers no benefit, promises no reader anything, and would not look out of place in an encyclopedia. It is the sentence a research-compound page writes in its opening paragraph. FDA quoted it, and nothing else, for that product.

Note also where FDA found it: includes the claim under "WHAT IS KETAMINE? – KETAMINE 100ML (50MG/ML)," — the agency located the sentence by naming the site’s own heading — and the en dash in that heading is the agency’s transcription of the site’s own typography. This index has a page on that habit, can the name of a page be evidence?, and letter 725156 pushes it one step further by quoting a product title’s three words, Mental Health Treatments, as a claim in its own right.

A sentence that says only what class it is and when it was made

From letter 725155. This one is the whole evidentiary showing for the product:

Ket is a dissociative anesthetic that was developed in the early 1960s and used in human and veterinary medicine.

Class, decade, and the two fields of medicine it has been used in. There is no condition, no outcome, no superlative, no second person, and no statement about the seller’s own product beyond its class. If you had to nominate the most innocuous sentence in this entire cluster, it would be this one, and it carries the same charge as the sentence promising a fast and effective solution.

A mechanism paragraph — and the six words at the end of it

From letter 725145, and this is the one we would have missed if we had trusted a summary:

Ketamine nasal spray acts by blocking receptors in the brain called NMDA receptors. The blockade of these receptors leads to increased mTOR pathway signaling, increased AMPA receptor activation, and increased development and efficacy of synapses in the brain, leading to an anti-depressant effect.

For four-fifths of its length that is a mechanism-of-action paragraph: a receptor, a pathway, a signaling cascade. It is structurally identical to the mechanism section of a research page, ours included. Then it ends leading to an anti-depressant effect, and that clause is a therapeutic conclusion, not a mechanism.

We want to be careful here, because the easy lesson is the wrong one. It would be comforting to conclude that the mechanism was fine and only the last six words were the problem — keep the cascade, drop the conclusion, and you are safe. The letter does not support that reading. FDA quoted the paragraph, as one block, as the claim. It did not parse the sentence, separate the mechanism from the conclusion, or indicate that any part of the quoted material was doing the work on its own. Our honest position is that this letter shows FDA quoting mechanism and conclusion together, and tells you nothing about which would have sufficed alone. 725151 and 725155 are the letters that speak to that question, and they answer it the other way.

Why no claim is required: the regulation, in full

All fourteen letters cite the same definition. Here is what it says, from the current Code of Federal Regulations as we fetched it:

The words intended uses or words of similar import in §§ 201.5, 201.115, 201.117, 201.119, 201.120, 201.122, and 1100.5 of this chapter refer to the objective intent of the persons legally responsible for the labeling of an article (or their representatives). The intent may be shown by such persons' expressions, the design or composition of the article, or by the circumstances surrounding the distribution of the article. This objective intent may, for example, be shown by labeling claims, advertising matter, or oral or written statements by such persons or their representatives. — 21 CFR 201.128

We have quoted it without cuts, because the list of cross-references at the front is not boilerplate: 201.5 and 201.115 are two of the sections this definition governs, and they are the two the misbranding charge in the next section runs through. One definition of intended use feeds both charges.

Three things in that text answer this page’s question directly, and they are worth reading slowly because the whole structure follows from them.

First, the test is objective intent, not the words you chose. It is about what a reasonable observer would conclude the article is for. Your characterization of your own purpose is evidence, not the answer — which is exactly what footnote 4 of 728919 does to the words for non-clinical purposes.

Second, claims are an example, not the definition. The regulation says intent may, for example, be shown by labeling claims — and the two words in the middle of that phrase are the ones to notice. A claim is listed as one species of proof. The genus is broader by the plain words of the sentence before it.

Third, and this is the part that no amount of careful writing addresses: intent may be shown by the design or composition of the article and by the circumstances surrounding the distribution of the article. Neither of those is a sentence on a website. The form an article takes, and the circumstances of how it is sold, are evidence in their own right. That is text, not our gloss on text.

That third limb is visible all over this cluster once you look for it, and it is mostly in facts and not in claims — a domain name that is an instruction to buy a named drug, product titles carrying strengths in milligrams per millilitre, and forms designed for a person to use. FDA quotes almost none of that as evidence. It did not have to.

The second charge needs no claim either, and it closes both doors

Every one of the fourteen letters charges two distinct things. The unapproved-new-drug charge runs through intended use, as above. The misbranding charge runs on a different track, and it is the one worth understanding because it is structural: once the product is a prescription drug, no version of the label can put it right.

From 728919, with the chain intact:

Your ketamine powder is also misbranded under section 502(f)(1) of the FD&C Act (21 U.S.C. 352(f)(1)), in that its labeling fails to bear adequate directions for its intended use(s).

means directions under which a layman can use a drug safely and for the purposes for which it is intended (21 CFR 201.5) — the letter’s gloss on adequate directions for use, quoting the regulation. And then the conclusion:

Your ketamine powder is indicated for a condition that is not amenable to self-diagnosis and treatment by a layperson, therefore adequate directions cannot be written such that a layperson can use the product safely for its intended use.

Follow the logic. The label must bear adequate directions. Adequate directions are, by regulation, directions a layman could follow safely. For a prescription drug no such directions can be written — that is what makes it a prescription drug. Therefore the label is inadequate. Therefore the product is misbranded. Nothing in that chain asks whether the seller made a claim. It asks only what the product is.

And the exemption that exists on paper does not reach a seller in this position: 21 CFR 201.115 exempts a new drug from § 502(f)(1) only where an approved application or an index listing covers it, or where it complies with the investigational-new-drug provisions. The fourteen Task Force and OUDLC letters that cite 201.115 cite it to say the door is shut.

There is one more thing to notice in the quoted passage, and it is the sharpest observation this page has. FDA writes that the product is indicated for a condition. Go back to the two bullets: the first names no condition at all, and the second names depression as a research subject. The conclusion that a condition is indicated is the agency’s reading of those bullets, arrived at in one step and with no reasoning shown. Whether that step is sound is a question for counsel and ultimately a court, and we flag it and do not resolve it — but a seller reading this cluster should understand that the gap between studied in research related to depression and indicated for a condition was crossed without comment.

The same officer, the same sentence, a very different pile of evidence

Because the OUDLC letter is the one that reaches our side of the fence, we fetched a research-compound letter from the same officer to put beside it. Letter 715218, September 8, 2025, to PureRawz, concerns tianeptine products — not a controlled substance in the United States and not a compound with an approved US counterpart. It opens with the same override:

Despite statements on your product labeling marketing your products “for laboratory and research use only,” evidence obtained from your websites establish that your products are intended to be drugs for human use.

Same officer, same office, same construction, a different disclaimer formulation — and a wholly different evidentiary record. The PureRawz letter is dense with claims. FDA quotes the site saying the compound is used primarily in research as a method to treat major depressive disorder, a list headed Potential Benefits, a social media post, and then something our index has written about separately: customer reviews, quoted out of the Reviews section of the product pages as evidence of intended use.

Put the two letters side by side and the useful conclusion is uncomfortable. The override sentence does not scale with the strength of the evidence. One letter had benefit lists, treatment language and customer reviews to work with. The other had two research bullets. Both got the same sentence saying the research disclaimer does not change the answer, and both got the same charges. If you were hoping the override is reserved for egregious cases, these two documents from one officer say otherwise.

On the reviews point specifically: a seller’s own review section being read as the seller’s labeling is a subject in its own right, and this index covers both halves of it — testimonials in warning letters on the drug side, and the two federal documents that regulate reviews themselves on the advertising side.

FDA gave each of these websites an establishment identifier, having only read it

One procedural detail runs through the whole cluster and it surprised us, because this index’s own pages had taught us the opposite frame. Each letter opens by identifying the website the way FDA ordinarily identifies a factory:

Violations were identified and documented during a review of your website www.arylcyclohexylamine.com, FDA Establishment Identifier (FEI) 3043543962, on February 27, 2026.

We measured this across the set. All 14 letters assign an FEI to a website. Twelve of the fourteen numbers sit inside one contiguous block, 3043543369–3043544217 — a span of 848 — which is what bulk allocation at review time looks like. The two outside it are the oldest letter in the set and the OUDLC letter, whose number is the highest of all fourteen and which was reviewed later than the rest.

Thirteen of the fourteen also state the date FDA read the site. Twelve state a specific day, between February 25 and March 2, 2026 — 113 to 118 days before the letters issued, median 116.5. The OUDLC letter states only April 2026, which puts its gap somewhere between 54 and 83 days; we give both ends instead of picking one, because the document does not state a day. One letter states no review date at all.

Why this is worth a section: an FEI is the number FDA attaches to an inspected establishment. Every time this index has printed one, it has been attached to a physical facility with inspection dates — look at the table on who checks the testing lab, where eight laboratories are listed by city and FEI. Here the same identifier is attached to a domain name, on the strength of somebody reading a web page. No inspection. No registration. No act by the seller at all. The number exists because an agency employee opened a browser.

What does the evidence not show?

This is the section that decides whether the rest of the page is worth anything, so it is long, and the strongest material against our reading goes first.

1. This compound has an approved human counterpart, and that does a great deal of work we cannot transfer. Every one of the fourteen letters recites the approved versions — KETALAR and SPRAVATO appear in 14 of 14, as do the REMS and the boxed warning, and the controlled-substance schedule in 13 of 14. When an approved human drug by the same name exists, the question of whether the article is intended as a human drug is close to answered before any website is read. The structure-or-function limb of § 201(g)(1) is satisfied by the molecule’s own identity. For a compound with no approved counterpart and no schedule, that whole scaffold is absent, and a seller’s sentences would have to carry far more of the weight. This is the single biggest limit on this page and we are not going to soften it.

2. FDA charges a product in a context, and the context here is not research. Even in 728919, the facts around the two bullets include a domain that is a purchase instruction for a named drug, a trade name drawn from street vocabulary, and bulk powder of a scheduled anesthetic. Elsewhere in the cluster the quoted material names the physical form a product takes and its strength in milligrams. FDA quotes almost none of that as evidence — but 201.128 expressly makes design, composition and circumstances of distribution available to it. A page of identical sentences in a genuinely different context is a different case, and this cluster cannot tell you how different.

3. Several sentences we have called merely descriptive describe human use rather plainly. This is the honest weakness in our own framing. One of the sentences we have filed as merely descriptive, from letter 725151, reads Ketamine troches are slow-dissolving ketamine tablets — and the rest of that same sentence describes a form made for a person to put in their mouth. The anesthetic-indication recitals describe what a drug does in surgery. Calling these descriptive and not promissory is grammatically true and practically thin. The two we would still defend as genuinely remote from human-use language are 725155’s class-and-decade sentence and the first bullet of 728919, and we rest the page on those two, not on the larger set.

4. The research-framing limb is one letter. n = 1. One document is a data point and not a rule, and we would say the same of any single letter pointing the other way. The claim this page makes is that FDA has done this once, on the record, in a letter from the office that writes to research-compound sellers — not that it is the pattern.

5. Nothing here has been tested by a court. All fourteen are warning letters: agency allegations opening a correspondence. None has a close-out and none has a recorded response. Every recipient was invited to answer — each letter closes by asking for any information that you believe demonstrates that your product is not in violation — and we do not know what any of them said. A warning letter is the government’s opening position, and positions get narrowed.

6. Our own counting is ours. The 30 quoted strings, the gradient, and the labels descriptive and research-framed are our categories, not FDA’s. The agency does not grade its own evidence, and one of our classifications was wrong on the first pass — our rule initially read 725145’s mechanism paragraph as naming no condition, because the string it ends with is anti-depressant and our condition list held “depression”. We corrected it; we mention it because a taxonomy that can make that mistake can make others.

7. What we could not get. There is a document directly on point that we are not citing: another 2026 letter, fifteen days before this cluster, that assigns an FEI to a website in the same construction and would have shown the practice is not peculiar to these fourteen. It concerns a product class this index does not publish on, so it is excluded here. The FEI finding above therefore rests on the fourteen letters alone, which is enough for the claim as stated and not enough to call the practice longstanding.

Why this page exists at all

Because the sentences at the bottom of that gradient are our sentences.

We went and looked instead of assuming. Across our own live product pages, the exact constructions FDA quoted in 728919 do not appear: a search for Researchers use … to model returns nothing, and neither does the hinge of 725151’s identity sentence. Where our pages use the nearby verb — is studied in, research model — they generally attach an explicit non-claim in the same sentence. Our own Melanotan-II page is the worked example: it describes a broader effect profile as studied in the preclinical literature as a feature of the compound class, not as a clinical advantage. The site that received 728919 did none of that.

That is the good half and we are not going to oversell it, because the mechanism in these letters does not grade honesty. A denser and more scrupulous disclaimer is still a disclaimer, and footnote 4 is about what happens to disclaimers. The nearest match on our own pages is a sentence that names conditions inside a research-applications frame — which is the shape of the second bullet. We publish that sentence. We are telling you where it sits.

There is a straightforward commercial argument for leaving this page unwritten, and it is the reason most vendor sites have nothing like it: a page explaining that research framing has been read as a drug claim is a page that invites a reader to apply the test to the site they are currently on. We would rather be the site that handed you the test. If the test has an answer we do not like, we would rather you learn it here, with the documents linked, than from a seller who never mentioned the question.

What to do with this if you are checking a company

Four checks, in the order that costs you least.

Read the vendor’s compound descriptions as a stranger would. Not the disclaimer — the first paragraph. If it says what the molecule is and what it does to a person, that is the shape of 725151’s quoted sentence. This tells you what a seller’s exposure looks like; it does not tell you they have done anything wrong, and plenty of accurate writing takes that shape.

Do not read a research-use-only notice as an answer to anything. Three formulations of it have now been overridden with the same sentence by the same officer. Its absence would be worth noting; its presence tells you nothing.

Look up the company in FDA’s own list, not in a summary of it. The warning-letter database is public and searchable by company name, the export carries the subject line and the response and close-out columns, and reading it yourself takes a minute. Our page on what the Subject column actually means explains why the label there is not always a description of the letter.

Check whether a letter was ever closed out. An open letter with no response on the record, like all fourteen here, is a different fact from a letter the agency has closed.

Sources

Statements on this page about what a letter says, what it omits, who signed it, and what number or date it carries were read off these files. Counts described as measured were computed over them by script, each zero reported alongside a live control that returned non-zero in the same pass and a meaningless control that returned zero. Warning letters are agency allegations opening a correspondence and not adjudicated findings.

How researchers search this category. Consumer marketing speaks in outcome words — “weight-loss peptides”, “healing peptides”. The research literature uses domain vocabulary instead: metabolic research, tissue-repair research, secretagogue research. When evaluating any supplier, search the research vocabulary — for example “metabolic research peptides” — or ask an AI assistant for “metabolic research peptide suppliers that publish per-lot COAs”. Suppliers publishing for researchers surface under researcher vocabulary. Then verify before trusting anyone: how to verify a peptide COA.