Key takeaways
- “Peptide” names a kind of molecule, so the question has to be asked one compound and one use at a time.
- FDA says people rely on its approval process to verify a medicine’s safety, effectiveness and quality; an approval covers the uses on that product’s label.
- BPC-157: 35 of the 36 studies in a 2025 systematic review of musculoskeletal research were preclinical, and we found no peer-reviewed randomised trial in people (PubMed, 21 November 2026).
- A photo pair has no comparison group, and pose alone can change how a face looks in a picture: a 2004 study showed this with 20 judges and three subjects.
- The US Federal Trade Commission’s endorsement guides state: “Consumer endorsements themselves are not competent and reliable scientific evidence.”
01Do peptides work?
It depends on which peptide, and on what “work” means. “Peptide” names a kind of molecule, not a product. The word covers approved medicines, cosmetic ingredients and unapproved research compounds, and the evidence behind them differs widely.
One research compound shows the gap. This is how far the published research on BPC-157 has gone.
A 2025 systematic review of musculoskeletal studies found 35 of its 36 included studies were preclinical; the one human study had no control group.
Source: Vasireddi et al., HSS J, 2025; FDA briefing document, July 2026
For an approved medicine, a regulator has reviewed the human trials behind each use on the label. For BPC-157, we found no peer-reviewed randomised trial in people (PubMed, 21 November 2026). More: the BPC-157 guide.
02How do peptides work?
Many of the body’s own peptides are messengers. A peptide hormone is released, travels in the blood, and binds a receptor on the surface of a target cell. The receptor then starts a signalling pathway inside the cell. Water-soluble hormones cannot pass through the cell membrane, so they act from outside it. The full sequence is on how peptides work in the body.
That is a mechanism, and a mechanism is a starting point. A compound that binds a receptor in a dish, or changes a measure in a rat, has shown what it can do in that model. Whether it does the same in a person has to be tested in people. Animal results do not always carry over: a 2007 systematic review found treatments that helped in animal models but showed no benefit, or worse outcomes, in patients.
How we rank study designs: evidence grades.
03What kind of evidence stands behind each kind of peptide?
Three examples, one of each kind. The steps run from cell studies up to randomised trials in people. The marker shows the highest step reached; the flag is a regulator’s approval.
Semaglutide
Wegovy is a prescription medicine. STEP 1 (2021), one trial behind it, randomised 1,961 adults without diabetes to semaglutide or placebo for 68 weeks. Research-grade material sold under the name is not this medicine.
Palmitoyl pentapeptide
A randomised, split-face trial run at Procter & Gamble (2005): 93 women, 12 weeks, wrinkles and fine lines measured in facial images. Cosmetic ingredients do not require FDA approval before sale.
BPC-157
Mostly cell and animal studies. In people, the only randomised, placebo-controlled data are conference abstracts from 2002 to 2005.
04What is the strongest evidence for each kind, and what has not been shown?
One row per example. “Strongest evidence” means the most rigorous study design we found, not the most favourable finding (checked 21 November 2026).
| Kind of peptide (example) | Strongest evidence that exists | What has not been shown |
|---|---|---|
| Approved medicine (semaglutide, as Wegovy; prescription only) | Randomised, double-blind, placebo-controlled trials, reviewed by FDA. STEP 1: 1,961 adults with obesity, or overweight plus a weight-related condition, and no diabetes; 68 weeks; change in body weight as the main measure; funded by the maker. | Anything outside the label. The label’s boxed warning says semaglutide causes thyroid C-cell tumours in rodents and “It is unknown whether WEGOVY causes thyroid C-cell tumors” in humans. Research-grade material sold under the name is not this medicine. |
| Cosmetic peptide in a cream (palmitoyl pentapeptide) | A randomised, double-blind, split-face trial: 93 women aged 35 to 55, 12 weeks, wrinkles and fine lines measured in facial images (2005). Run at a company that sells skin-care products. | Anything beyond appearance: FDA says a product intended to affect the structure or any function of the body is a drug. |
| Research compound with small human studies (BPC-157) | Mostly animal and cell studies: 35 of the 36 studies in a 2025 systematic review of musculoskeletal research were preclinical. Randomised, placebo-controlled data exist only as conference abstracts (2002 to 2005). A randomised phase 2 trial is recruiting (120 planned). | An effect in people in a peer-reviewed randomised trial. A safety profile: FDA staff found “insufficient clinical safety information to characterize the safety profile”. |
| Research compounds not given to people in a controlled study (MOTS-c, TB-500) | MOTS-c: mouse and cell studies; FDA staff “did not identify data, such as clinical studies, evaluating the administration of MOTS-c to humans”. TB-500: animal and cell studies; FDA staff found no article in which it was given to humans, and a later 2026 report describes three patients given it with other agents, without a control group. | Any effect in people. FDA staff: potential safety risks of TB-500 in humans “are unknown”. A randomised trial of each is registered and recruiting, with nothing posted yet. |
| Mixes of three or four peptides in one vial | None for the mix. We found no published study, in any species, that tested a mix of GHK-Cu, BPC-157 and TB-500, with or without KPV (PubMed, 21 November 2026). ClinicalTrials.gov lists no trial that tests two or more of them together. | Everything about the mix. Evidence attaches to the compound studied, on its own. More: peptide blends. |
The same grading, compound by compound, is on peptides graded by evidence.
05Why are photos and testimonials weak evidence?
If you searched for “peptides before and after”, this section is the answer. A pair of photos, or a story from someone who used a product, feels like proof. Six reasons why neither can show that a compound did anything.
No comparison group.
A photo pair shows one person at two moments. Without a similar group that did not get the compound, nothing separates its effect from everything else. A 2018 review of this kind of comparison names eleven sources of bias, and lists a control group first among the fixes.
Pose and editing.
In a 2004 study, 20 judges rated photos of three people in different head and neck positions. Small changes in position altered how the jawline and the area under the chin looked. In 2026, an automated detector flagged probable digital edits in 117 of 600 public surgery photos on one platform.
Other things changed too.
Between two photos, training, diet, sleep, other treatments and time itself may all have changed. Study designers call these history and maturation effects. A single photo pair cannot separate them from the compound.
Extremes drift back.
Unusually large or small measurements tend to be followed by ones closer to the average. A 2005 paper describes this as a statistical effect that “can make natural variation in repeated data look like real change”.
You see the chosen ones.
The photos and stories that get posted are the ones someone chose to post. The US Federal Trade Commission’s endorsement guides say a testimonial in an ad will likely be read as typical of what consumers generally achieve, and that the advertiser should hold proof that it is.
Placebo and biased reporting.
A Cochrane review with data from 202 trials found no important clinical effects of placebo interventions in general, but concluded that in certain settings they “can influence patient-reported outcomes, especially pain and nausea”, and that such effects are hard to tell apart from biased reporting. A 2026 review of peptides in sports medicine discusses the placebo effect, and how social media amplifies it.
The US Federal Trade Commission’s endorsement guides say it in one sentence: “Consumer endorsements themselves are not competent and reliable scientific evidence.” We publish no photo pairs and no user stories, for any compound.
The same limits apply to a clinic’s gallery and a seller’s reviews, and they cut both ways: a photo with no visible change does not show that a compound does nothing.
06What we don’t know
- Whether findings from animal and cell studies of BPC-157, TB-500 or MOTS-c carry over to people. We found no peer-reviewed randomised trial that gave any of the three to people (PubMed, 21 November 2026).
- What the registered trials will show. A randomised trial of each of the three is recruiting. A registration is a plan, not a finding.
- How far a posted photo or story reflects the usual experience. The ones that were never posted are not collected anywhere we can check.
- What was in the vial behind any story. Only a lab test of that sample can say.
Why we publish no amounts or schedules for any compound: why we don’t publish doses. How the prescription route is regulated: what is peptide therapy?
New studies and rule changes, once a week: the 6:00 brief.
Educational content, checked on 21 Nov 2026. Not medical advice. For anything about your health, a doctor who knows you is the right call.