Collagen improves your skin

About 50 randomized trials of collagen peptides taken by mouth for skin have been published, nearly all against a placebo, and a further search found about 20 more small controlled trials, about half testing collagen mixed with other ingredients. The largest pooled analysis, Batool 2026, combined 35 trials with 2,534 participants and found average improvements in instrument-measured skin hydration and elasticity and in water loss through the skin. It found no effect on wrinkles once one outlying trial was set aside, and an apparent change in the density of the deeper skin layer did not hold up under its further checks. The trials are small: the 35 averaged about 72 participants each.

An earlier analysis, Myung and Park 2025, split 23 trials by funding and by study quality. Trials funded by pharmaceutical companies showed improvements, while trials without that funding showed no effect on hydration, elasticity or wrinkles. Trials rated high quality showed no effect in any category, while lower-quality trials showed an improvement in elasticity. Batool 2026 did not repeat that split. It did report signs that small trials with no effect are missing from the published record, which would make the pooled improvements look somewhat larger than they are. A 2026 re-analysis, Grimes and Mahto, went back to the trial reports behind two earlier pooled analyses, one of them Pu 2023, and found errors in both, including standard errors entered as standard deviations and readings from different instruments combined as if they measured the same thing. With the data corrected, it found no link between the dose of collagen and the difference from placebo across 16 hydration and 18 elasticity trials, and its authors concluded that the trials do not show a benefit. It did not report a corrected average difference from placebo, so its conclusion rests on the errors it found and the absence of a dose link, and it did not re-check Batool 2026.

Across the 35 pooled trials, doses ran from 500 mg to 10 g a day and most lasted 12 weeks. Batool 2026 found that how long collagen was taken mattered more than the dose: hydration readings were better as early as week 4, while elasticity and water loss improved only at around 12 weeks. Its hydration and elasticity results are reported after outlying trials were removed, and before that removal the elasticity analysis showed no benefit. An earlier meta-analysis of 26 trials, Pu 2023, also found improved hydration and elasticity, with the hydration effect varying with the collagen source and with how long it was taken; it is one of the two analyses in which the 2026 re-analysis found data errors.

In three of the trials cited here, the product was a proprietary collagen peptide preparation made or supplied by a commercial partner, taken for 12 weeks at 1 g to 10 g a day. Kim 2018 gave 64 women aged 40 to 60 with sun-damaged skin 1,000 mg a day of a fish-derived peptide. Hydration and several wrinkle measures improved compared with placebo, and two of three elasticity measures were higher at 12 weeks. The trial ran in 2012 and was published in 2018, and two of its authors worked for the company that supplied the peptide. Wang 2025 gave women 5 g a day, with 77 analysed. Skin moisture, water loss, firmness and dermal density were better than placebo and stayed that way four weeks after stopping, while a standard elasticity measure and the thickness of the outer skin layer did not differ. That trial was sponsored by the manufacturer and registered after it had finished.

Paula-Vieira 2026 randomly assigned 119 women aged 35 to 55 to 2.5 g or 10 g a day or to a control group that took nothing, with no placebo. After 12 weeks, elasticity and hydration had risen and wrinkle length had fallen in both collagen groups, and wrinkle count had fallen in the 10 g group, with no change in the control group. The paper reports these as changes from each group's own starting point. Public research foundations funded the study, the manufacturer supplied the peptide and had paid the lead investigator consulting fees, and the trial was registered after its data were collected. The largest placebo-controlled trial published after the pooled analyses, Wisutthathum 2026, gave 140 Thai adults aged 30 to 60, about three-quarters of them women, a daily drink of 5 g of hydrolysed collagen with glutathione and vitamin E, or a placebo drink, for 12 weeks; 135 finished. Skin moisture rose in both groups with no difference between them. Two elasticity readings were higher than with placebo at week 4 but not at weeks 8 or 12, one of four measures of wrinkle roughness was lower at week 12, on a measure where the collagen group had started lower, and dermatologists' wrinkle grades did not differ. The supplement company that supplied the drink funded the trial, and it was registered after it had finished.

Collagen is broken down in digestion, but some short fragments get through. In healthy volunteers, a collagen fragment called Pro-Hyp and related peptides rose in the blood and peaked one to two hours after a dose (Iwai 2005). In mouse skin cells, these fragments trigger growth of fibroblasts, the cells that make collagen. Asai 2019 found that earlier lab results were confounded because standard cell-culture serum already contains the same peptides, and the effect appeared once they were removed. Whether blood levels after a supplement dose do this in human skin is not shown in these sources.

Kim 2018 reported no adverse symptoms, Wang 2025 reported no adverse reactions, and Wisutthathum 2026 reported no adverse events linked to the product and normal blood and urine tests. Kim 2018 enrolled 64 women and Wang 2025 analysed 77. The trials measured skin with instruments (moisture meters, suction devices, ultrasound and image analysis), most over 8 to 12 weeks. Batool 2026 noted that this may be too short to show a change in wrinkles. The sources here do not establish whether changes of this size are visible, or how long they last beyond a four-week stop. Three of the four trials cited here enrolled only women, and the fourth was about three-quarters women.