"Peptides or collagen" is a common question on the skincare shelf. It mixes two questions together: what can work on the skin from outside, and what can be swallowed. This page separates them and counts the evidence for each.
The size problem
Collagen is enormous. UniProt lists the human collagen alpha-1(I) chain (entry P02452) as a precursor of 1,464 amino acids and 138,911 daltons. A type I collagen molecule twists three such chains together into a rope-like triple helix.
The usual reference point for skin penetration is the 500-dalton rule of Bos and Meinardi (Exp Dermatol 2000, PMID 10839713): compounds much above that weight generally do not pass through intact outer skin on their own. One collagen chain is about 280 times that line. It is a rough rule, not a law, but at that ratio there is no argument about the direction.
Cosmetic peptides sit near the line, not far above it. Acetyl tetrapeptide-5 is 492.5 daltons and acetyl tetrapeptide-2 565.6. Palmitoylated peptides such as Matrixyl are heavier, and the fatty tail exists to help them in. Whether any of them reaches living cells in useful amounts is still debated, and peptides for skin covers that argument. The difference in scale is still the starting point.
So a collagen cream is a film. Large proteins hold water at the surface and can make skin feel smoother. That is a real, short-term moisturising effect. It is not a way of adding collagen to the skin.
The naming problem
"Collagen" on a label can mean three different things, and searches mix them up:
| Label wording | What it is | Size |
|---|---|---|
| Soluble collagen, collagen | Intact or near-intact protein | Hundreds of thousands of daltons |
| Hydrolyzed collagen, collagen peptides | A mixture of fragments made by breaking collagen down | Variable; mostly thousands of daltons |
| Tripeptide-29 (collagen tripeptide) | One defined fragment, Gly-Pro-Hyp | 285 daltons |
The third is under the line. Tripeptide-29 has its own entry, which found that five of its six randomized trials gave it by mouth. Front labels blur the line further: one mass-market serum calls palmitoyl pentapeptide-4 a "collagen peptide", as our Olay Super Serum analysis shows.
Hydrolyzed collagen is technically peptides. That is why "collagen peptides vs peptides" is a muddled question: the difference that matters is between a random mixture of fragments, sold mostly as a drink, and a single synthetic sequence designed to signal cells.
The topical evidence: collagen is usually the outcome, not the ingredient
PubMed, 5 October 2026: randomized controlled trials with collagen in the title, applied topically or as a cream, with oral use and supplements excluded. 10 records.
Read by title and abstract, most are not about collagen as an ingredient at all. They are trials of something else, such as estradiol cream, a fullerene complex or a wound dressing, that measured collagen in the skin as the result. Three test a collagen-containing product:
- Recombinant human collagen serum on stretch marks. Li and colleagues, African Journal of Reproductive Health 2026 (PMID 42454417). Twenty postpartum women applied the serum twice daily to one randomly chosen side of the abdomen for 4 weeks, with the other side untreated. The treated side improved more on elasticity, striae area and roughness. The control side received nothing, not a base serum, so the effect cannot be separated from moisturising.
- Recombinant type III collagen by injection plus a "multi-peptide" serum. Qin and colleagues, Journal of Cosmetic Dermatology 2026 (PMID 42087486), a split-face study. Injection puts the protein under the barrier, so it does not answer the cream question.
- A collagen tripeptide with a sugar compound in sensitive atopic skin, from 2009, which the tripeptide-29 entry reads in full.
That is a thin topical record for an ingredient sold on many labels.
The oral evidence is a different and larger question
The same type of search for oral collagen peptides or hydrolysate with a skin endpoint returned 38 randomized trials. This site reads them on collagen peptides for hair and skin, including two meta-analyses that point the same way and many trials funded by the collagen's maker.
A drink is digested. Its fragments reach the skin through the blood, if they reach it, which is a different route with different evidence. A good result for a collagen drink says nothing about a collagen cream, and the reverse is also true.
Where peptides came from: a fragment of collagen's precursor
Some of the best-known cosmetic peptides were designed from collagen itself.
In 1993 Katayama and colleagues cut a fragment of the type I procollagen C-terminal propeptide, the piece trimmed off when collagen is assembled, down to the smallest part that still worked. They reported Lys-Thr-Thr-Lys-Ser (residues 212-216) as the minimum sequence that strongly stimulated collagen and fibronectin production in cultured cells (J Biol Chem, PMID 8486721). Add a palmitic acid tail and that is palmitoyl pentapeptide-4, sold as Matrixyl.
When we read the matching position in UniProt's current human reference sequence, it says something slightly different. Residues 1430-1434 of collagen alpha-1(I), which are residues 212-216 of its C-terminal propeptide, read Lys-Thr-Thr-Lys-Thr, ending in threonine rather than serine. The chicken sequence has the same motif. The 1993 abstract does not state which species' sequence it used, and an older sequence reading or a different species could explain the difference. It does not change what the 1993 cell experiments found. It does show that the cosmetic peptide is a laboratory-defined fragment, not simply a piece of a person's own collagen.
What this supports saying
- Whole collagen is far too large to pass through intact skin; a collagen cream acts at the surface.
- Hydrolyzed collagen is a peptide mixture, mostly studied as a drink, with a much larger trial record than any topical collagen.
- Topical randomized evidence for collagen as an ingredient is thin; most collagen trials measure collagen as an outcome.
- Some cosmetic peptides were designed from collagen's precursor, and their own evidence is assessed separately on peptides for wrinkles.
- No trial compares a peptide product with a collagen product.
Sources
- UniProt P02452 (collagen alpha-1(I), human; length 1,464, mass 138,911 Da; C-terminal propeptide 1219-1464) and P02457 (chicken), sequences read 2026-10-05.
- Bos JD, Meinardi MM, Exp Dermatol 2000;9(3):165-9 (PMID 10839713), abstract read 2026-10-05.
- Katayama K et al., J Biol Chem 1993;268(14):9941-4 (PMID 8486721), abstract read 2026-10-05.
- Li C et al., Afr J Reprod Health 2026;30(13):86-100 (PMID 42454417), abstract read 2026-10-05.
- Qin X et al., J Cosmet Dermatol 2026;25(5):e70857 (PMID 42087486), abstract read 2026-10-05.
- PubMed E-utilities searches as described, run 2026-10-05.