What I check when I read a skincare claim.
Your test data, translated into claims that pass overseas review and still sell.
Amisha · Cell biologist · OsakaMEng Biomedical Engineering, University of Tokyo
The sort
Every skincare claim is one of four kinds, and each kind fails in its own way.
Composition claims
“Contains ten billion vesicles.”
Fail on measurement. Counted how, by whom, and does the number on the box match the number on the science page?
Mechanism claims
“Penetrates to signal your fibroblasts.”
Fail on delivery. The biology may be real in a dish; the sentence asserts it happened through intact skin.
Outcome claims
“Reduces wrinkles by 20%.”
Fail on study design. Who was measured, for how long, against what, and by instrument or by eye?
Status claims
“Clinically proven.” “Clinic grade.”
Fail on definition. The words have meanings to a regulator, and they are narrower than the ones marketing uses.
The point of the sort: it tells you where to look before you argue.
The four questions
Then I trace every number to its source and ask, in order:
- Was it measured in a dish or in a person?A percentage from fibroblasts in culture is not a percentage in a face.
- Was it the ingredient or the finished product?Testing the raw material is not testing the cream.
- What was the sample size, and was there a control?Thirty people with no comparison group cannot carry a number, and a reviewer will say so.
- Who paid?Sponsored is not disqualifying, but it changes how you read a positive result.
The barrier pass
What this looks like on a real page
Part A · A public case
“Clinically proven to lift the skin by 20%.”
On 13 May 2026 the UK Advertising Standards Authority upheld a complaint against an exosome face treatment sold with that sentence. The evidence behind it: thirty subjects, twenty-eight days, no placebo or control group, and a lifting score graded by eye on a ten-point scale rather than measured by instrument. The objective measures in the same study, roughness and firmness, did not show the lift. Two consumer studies were self-assessed, with no control. The ruling found the claims “had not been adequately substantiated, and exaggerated the effects.”
All four questions fail at once: person, yes, but a subjective endpoint; finished product, yes, but thirty with no control; and the brand paid.
What matters is what the brand did next. The 20% is gone. “Clinically proven” became “clinically tested.” “Regenerate” became “support.” The sample size is now on the page: sixty-one participants, four weeks, “90% agree skin appears firmer.” The product is still on sale, and the copy is better.
ASA ruling, 111 Skin Ltd, A25-1319497, 13 May 2026. Product page wording checked 14 September 2026.Part B · An anonymised sample
Three claims from one product page, and the three rewrites I sent back.
Anonymised. Brand, product and ingredient names removed; figures altered so the client cannot be identified. The pattern is the point: same data, accurate framing, still sells.
| What the page said | What I sent back |
|---|---|
| “Clinically proven. Eight times more powerful than vitamin C. Boosts collagen.” | “In laboratory testing on human skin cells, our active neutralised oxidative stress about eight times more potently than vitamin C and increased the expression of collagen genes.” |
| WhyThe paper behind the claim had no human subjects and no wrinkle endpoint; the eight-fold figure is a culture-dish potency, and “clinically proven” is the exact phrase a regulator upheld against in May 2026. | |
| “Clinical results in two weeks: 31.7% lifting improvement, 12.4% skin density improvement, 21.9% reduction in wrinkle depth.” | “In a two-week instrumental study of 24 women aged 35 to 55, wrinkle depth around the eyes decreased by an average of 22% and skin density increased by 12%. Open-label, no control group.” |
| WhyNo source, sample size or method was disclosed, and two-decimal precision reads as false precision to a reviewer. Disclosing the design turns the number from a liability into a credibility signal. | |
| “Stimulates cell regeneration and promotes healing and skin repair.” | “A purified active with a well-studied biology, in a daily formula made to support the skin barrier and the look of smoother, firmer skin.” |
| Why“Healing” and “repair” are the words a regulator uses to separate a drug from a cosmetic, and the data behind them belonged to an injectable, not to a leave-on serum. Appearance language keeps the product a cosmetic and keeps the clinic halo intact. | |
What the evidence does support
The science is rarely the problem.
- Extracellular vesicle cargo does change fibroblast behaviour in a dish.
- Photobiomodulation has a documented mitochondrial mechanism.
- Senescent cells are a documented part of skin ageing.
Delivery, product quality, and the distance between the study and the sentence are the problem. Saying so, out loud and every time, is what separates a reviewer from a cynic.
Send one product page
You get three claims and three rewrites back within the week. Scope is scientific communications support, not legal or regulatory approval, which stays with your counsel. It is the review your counsel is not equipped to do, because the question is biological before it is legal.
First read: three claims, three rewrites, one week. US$450, credited in full against a Claims Language Review (from US$1,500) if you continue.
I provide scientific communications support: evidence review, claims-language risk flagging, and ingredient rationale writing. I am not a physician, a regulatory professional, or a cosmetic chemist, and I do not conduct clinical testing. Final legal and regulatory approval of any claim should remain with your counsel or a qualified regulatory consultant. I work under confidentiality and IP-protection terms as standard.