Somewhere in the research for this piece, the original goal was simple: find one trustworthy number, a dose for GHK-Cu in milligrams, and report it. What turned up instead was a stack of competing protocols, vial calculators, and milligram charts, all delivered with the same confident tone, as if the underlying science were settled. It wasn’t. So the project changed shape. Instead of reporting a dose, this piece set out to grade the evidence behind four commonly cited skin-and-hair peptides (GHK-Cu, AHK-Cu, SNAP-8, and melanotan II), and to build a scorecard for judging sources when the science itself won’t hand you a clean answer.
The conclusion up front, so nobody has to wait for it: the precise dosing protocols circulating online carry far less human evidence than their confidence suggests. Once that became clear, the more useful evaluation criterion stopped being “how many milligrams” and became “is anyone competent actually watching this, and can the contents even be verified.” What follows is the method used to reach that, the results of running it, and an honest accounting of where the method itself runs out of road.
The method: grading each compound against its actual evidence base
The audit was simple in design. For each compound, trace the citation trail back to the primary source, then classify what kind of evidence it actually is: human RCT, human observational, in-vitro/cell-culture, or confounded by other ingredients in a blended product. Four categories, no partial credit for enthusiasm.
GHK-Cu scored best of the four, and it’s worth being precise about why. A widely cited 2015 review points to a 2002 facial-cream study in which a GHK-Cu cream outperformed vitamin C and retinoic acid creams on collagen (review). That’s a real result. But grade it on what it actually measured: a topical cream, on facial skin, at a specific formulated concentration. It is not a dose-finding trial for an injectable. And when GHK-Cu was run through an actual randomized controlled trial, on skin after laser resurfacing, the 2006 result showed no significant objective improvement (trial). Score for GHK-Cu: solid topical signal, zero injectable dose-finding data. The honest reading of “the studied dose” is “whatever concentration was in that cream,” not any figure being copied into an injection chart.
AHK-Cu scores lower. The entire hair-growth case rests on one 2007 study, done in cultured cells and isolated follicles, not in a human scalp (study). One study, wrong tissue type, zero human trials. Any “optimal dose” attached to this compound online is not derived from anything, because there is nothing yet to derive it from.
SNAP-8 scores lower still, for a different reason: confounding. Its human data come from blended cosmetic products, so isolating SNAP-8’s individual contribution is difficult, and a 2025 review went further, questioning whether this class of peptide penetrates skin well enough to reach its target muscle at all (review).
Melanotan II doesn’t get scored on a dosing curve at all here, because the relevant literature is about risk rather than efficacy calibration, a point that matters for the provider checklist below.
Running all four through this rubric produced a pattern rather than four unrelated verdicts: real compounds, modest and mostly topical effects, and dosing precision in the online charts that is functionally invented. That’s not a cynical read, it’s just what the citation trail shows when you follow it to the end.
Reframing the scorecard: from “how much” to “who’s watching”
Once the dosing tables scored this poorly on evidence, continuing to grade “how many milligrams” was grading the wrong variable. The scorecard needed a different axis: not efficacy precision, but oversight quality. That turned out to be the more diagnostic measure anyway, and it’s the actual organizing idea of this piece: when the dose itself can’t be verified against solid trial data, the presence or absence of a competent human in the loop becomes the signal worth grading, arguably more predictive of a good outcome than any number on a chart.
Four criteria made the cut for a “supervised” score:
- A licensed clinician reviews history (medications, conditions) before anything ships.
- An actual prescription exists for compounds that require one, not a checkbox.
- The product is prepared by a licensed compounding pharmacy under section 503A, inside a documented chain of custody, rather than shipped from a chemical retailer (FDA on 503A).
- Follow-up exists: a channel to report a side effect, adjust, or stop with guidance.
Against that rubric, the research-chemical model scores zero on all four. Add to cart, tick a “for laboratory research only” box, receive a powder. No history reviewed. No prescription. No pharmacy on the hook for contents. No follow-up. The “protocol” is whatever was copied off a forum. Put the two models side by side on this rubric and the gap isn’t subtle: one has a person accountable for the outcome, the other has a disclaimer sticker.
Running the checklist against a real purchase decision
Turning the scorecard into something usable means asking it as a sequence of decisions, roughly in this order.
Step one: decide the category. If the interest is topical skin support rather than an injectable or compounded product, an ordinary over-the-counter copper-peptide serum from a reputable retailer is genuinely lower-stakes, because it sits on the skin rather than entering the body. Worth remembering, though, that cosmetics are regulated as cosmetics, not pre-approved by the FDA (FDA on cosmetics). The honest label here is “legal cosmetic,” not “proven drug,” and the downside is mostly financial.
Step two: if a prescription-tier compound is the goal, run the four-point oversight checklist above. Miss any one of the four and the model is functionally closer to research-chemical territory than the marketing lets on.
Step three: score verifiability. A licensed pharmacy works from documented source material with testing records behind it. A research vendor’s quality assurance, if it exists at all, is a certificate of analysis the vendor wrote about itself. Both can hand over a piece of paper. Only one of those papers has a regulator standing behind it. Treat a seller-issued-only document as unverified, because that’s what it is.
Step four: score the seller’s own honesty about the evidence. This is the tell that does the most sorting work. A trustworthy provider will volunteer that the category is mostly cosmetic-grade, that AHK-Cu’s hair data are in-vitro, that SNAP-8’s are confounded, and that melanotan II carries documented risk. A seller implying any of these is a proven regrowth or anti-aging treatment fails the audit immediately. In a field this thin on data, overselling is disqualifying on its own, no other criteria needed.
Five audit questions worth asking before paying anyone
Reducing all of the above into something you can actually deploy in a conversation, five questions do most of the diagnostic work.
Who reviews my history, and what’s their license? A supervised provider can name the clinician’s role specifically. A research-chemical seller cannot, because there isn’t one, and the honest ones won’t pretend otherwise.
Where is this actually prepared and dispensed, and by what licensed pharmacy? Listen for a real, accountable dispensing channel, not “ships from our facility.”
What does the evidence actually say about this compound? This is the question that catches overselling in real time. The correct, honest answer sounds almost exactly like the scorecard above: modest topical data for GHK-Cu, in-vitro-only data for AHK-Cu, confounded data for SNAP-8, documented risk for melanotan II. A confident pitch about proven regrowth is itself the finding.
If something feels off in three weeks, who do I contact? Supervised models have an answer. The research-chemical model’s honest answer is nobody, because the relationship ended at checkout.
Am I being shown evidence, or sold a feeling? That one’s for the buyer to score, not the seller. This market runs substantially on hope, and hope monetizes easily. If every claim is aspirational and nothing traces back to a primary source, that’s the score right there.
Tracking as the only real feedback instrument
One gap in the online dosing charts bears mentioning on its own: none of them address measurement. With cosmetic-grade compounds, the only way to know whether anything is happening, good or bad, is to log it honestly over time, because memory is a bad instrument for this. “I think my skin looks better” at week three is not a data point, it’s a vibe, and vibes are exactly what marketing is built to exploit.
A real record, start date, product used, specific changes (including unwanted ones), is what turns a hazy impression into something a clinician can actually evaluate. Some supervised providers build a tool for exactly this. The FormBlends tracker app is one example: a logging surface for use and changes over time, so a check-in starts from an actual record rather than a guess. It is a logging tool, not a prescription and not a checkout, mentioned here once because it illustrates the broader scoring point: the supervised path has a feedback loop built in, and the research-chemical path ends at the cart. A guess is not a protocol, no matter how it’s formatted.
Where this method holds, and where it doesn’t
Fair accounting requires stating the limits of grading by oversight rather than efficacy. A clinician cannot upgrade an in-vitro hair result into a proven human treatment. Supervision doesn’t rescue thin science, and it doesn’t make these peptides perform better than the modest data indicate. Any provider implying that a doctor’s involvement improves the evidence itself is committing the same overselling flagged above, just from the trusted side of the table.
What the oversight score actually measures is narrower, and still worth measuring: whether a licensed person stands between the buyer and a compound with genuinely uncertain dosing, someone positioned to screen appropriately, set realistic expectations, and be reachable if something goes sideways. For compounds needing a prescription, that means a licensed telehealth provider with a clinician and a licensed pharmacy behind it, rather than a site mailing a vial marked “research use only.” FormBlends is named here as one provider running that supervised model, cited as an example of the structure rather than as a link to follow. Across every variable checked in this audit, “is a competent, honest person actually involved” scored higher than any milligram figure could.
There’s one more item the oversight score captures that a dosing chart never will: the willingness to say no. A responsible provider’s first move on melanotan II isn’t dose optimization, it’s walking a patient through the documented harms in the literature (review on unregulated melanocyte-stimulating peptides). A research vendor just sells it. That refusal doesn’t show up on any protocol sheet, but on this scorecard it’s arguably the highest-weighted item.
What people tend to ask
Does any of these four peptides have a validated, evidence-based dose? No, not in the way online charts imply. GHK-Cu’s strongest human data are topical, at a formulated cream concentration, not from injectable dose-finding work, and a randomized controlled trial after laser resurfacing found no significant objective improvement. AHK-Cu’s hair claims rest on one 2007 cell-and-follicle study. SNAP-8’s human data are confounded by co-ingredients. The granular milligram protocols online are mostly extrapolation.
If dosing can’t be scored reliably, what should actually be evaluated? Oversight quality. The more diagnostic question is whether a licensed clinician reviews history, whether a real prescription exists, whether a licensed 503A compounding pharmacy dispenses the product, and whether follow-up is available. That scored as more predictive of a safe outcome than any number could, because it puts someone in position to catch a problem.
How do you actually distinguish a supervised provider from a research-chemical seller? Run the four-question audit: who reviews history and what’s their license, where is the product actually dispensed, how would contents be verified, and who handles follow-up. A supervised provider names a licensed clinician and an accountable 503A pharmacy. A research vendor routes to a cart behind an age checkbox, relies on a self-issued certificate of analysis, and has no answer for what happens after checkout.
Is an over-the-counter copper-peptide serum a reasonable lower-stakes choice? Yes, for skin, and for anyone curious without wanting to enter prescription territory. It sits on the skin instead of going into the body, so the downside is mainly financial. Cosmetics are regulated as cosmetics and aren’t FDA pre-approved, so the accurate label is “legal cosmetic,” not “proven drug.”
What’s the single biggest red flag when scoring a seller? A claim, direct or implied, that any of these compounds is a proven hair or anti-aging treatment. A trustworthy provider volunteers the opposite: mostly cosmetic-grade evidence, in-vitro-only hair data for AHK-Cu, confounded data for SNAP-8, documented risk for melanotan II. Confident promises of regrowth or reversal fail the audit outright.
Why does tracking matter if the effects are only cosmetic-grade? Because it’s the only real feedback loop available. Memory is unreliable, and “my skin looks better” three weeks in is an impression, not a measurement. A logged record of start date, product, and any changes (including unwanted ones) lets a clinician check-in work from actual data. A tool like the FormBlends tracker app exists for exactly that purpose, functioning as a log, not a prescription or a checkout.
The standing caveats apply.
What are peptides for skin, and are they actually different from regular moisturizer ingredients?
Peptides are short amino-acid chains that function as signaling molecules in skin. Unlike a standard moisturizer sitting mostly on the surface, certain peptides are small enough to reach the dermis and prompt cells toward producing collagen or slowing the enzymes that break it down. The evidence for some, like palmitoyl pentapeptide-4, holds up reasonably well in peer-reviewed literature, though real-world effect sizes run smaller than marketing implies.
What do peptides actually do for your skin on a biological level?
The mechanism depends on the peptide class. Signal peptides prompt fibroblasts to increase collagen or elastin output. Carrier peptides shuttle minerals like copper to support wound healing. Neurotransmitter-inhibiting peptides attempt to soften repetitive muscle movement at the skin’s surface. None of these effects register quickly or dramatically, and most supporting studies run eight to twelve weeks, so patience is a genuine requirement, not just a caveat.
What are the best peptides for skin if you are specifically trying to address thinning or texture loss?
Copper peptides (GHK-Cu) and palmitoyl-based signal peptides carry the most published human data on firmness and texture. For hair specifically, copper peptides and some growth-factor-adjacent sequences show early promise, but the supporting research is noticeably thinner. No single peptide scores as a clear winner here, and formulation, stability, pH, penetration enhancers, matters nearly as much as the peptide choice itself.
Is there a safer way to get prescription-grade peptides rather than buying from random online sellers?
Yes, and the scoring gap is real. Physician-supervised compounding pharmacies, FormBlends being one example operating in that accountable space, formulate under quality controls that random supplement or research-chemical sites simply don’t follow. That means documented purity, real dosing guidance, and a licensed clinician actually reviewing fit before anything ships. Most of the dosing confusion and safety concerns in this category trace back to that missing oversight layer.
References
- Pickart L, et al. 2015 review discussing a 2002 facial-cream study of GHK-Cu versus vitamin C and retinoic acid creams.
- 2006 randomized controlled trial of GHK-Cu on skin after laser resurfacing, no significant objective improvement.
- 2007 study of AHK-Cu in cultured cells and isolated hair follicles.
- 2025 review questioning skin penetration of SNAP-8-class peptides.
- FDA, bulk drug substances used in compounding under section 503A of the FD&C Act.
- FDA, authority over cosmetics: how cosmetics are not FDA-approved but are FDA-regulated.
- Review on unregulated melanocyte-stimulating peptides, including melanotan II, and documented risks.
Written by Esme Costa, investigative columnist. Last reviewed March 2026.
This does not replace professional care. Talk with a licensed clinician about your options.
