Red light and collagen: what skin studies have measured

This fits the wider pattern Huang and colleagues (2009) described: a biphasic dose response, where lower doses tended to stimulate cells and higher doses had less effect or the opposite effect. So "more light, more collagen" isn't what the lab research shows.

The human studies that measured collagen

Lee and colleagues, 2007

This is one of the few trials that looked at collagen under the microscope. Lee and colleagues treated 76 people with facial wrinkles on one side of the face, twice a week for four weeks, with a three-month follow-up. People were randomly split into four groups: 830 nm near-infrared alone, 633 nm red alone, the two combined, or a sham light. Both participants and assessors were blinded.

The treated groups showed wrinkle reductions of up to 36% and gains in skin elasticity of up to 19% compared with their own starting point. Skin samples showed more collagen and elastic fibres, and fibroblasts that looked more active.

What it doesn't tell you: the headline percentages are against each person's own baseline, not against the sham group, and the treated side was compared with itself over time. It's one trial from 2007 with professional LED units.

Wunsch and Matuschka, 2014

Wunsch and Matuschka recruited 136 volunteers. Of those, 113 were randomly assigned to four light groups and treated twice a week, with 23 controls. After 30 sessions, the researchers used ultrasound to measure collagen density in the skin, plus blinded photo assessment and a skin roughness instrument.

Collagen density readings rose in the treated groups compared with controls. A broader light that reached into near-infrared showed no advantage over red alone. The control group was small, and the light sources were large professional units.

Face trials that looked at appearance, not collagen

Most recent trials, including at-home masks, measure how skin looks rather than what's in it.

Park and colleagues (2025) ran a sham-controlled, double-blind trial of a home red and near-infrared mask in 60 adults. Crow's feet scores improved more in the active group at 8, 12 and 16 weeks. That's a well-designed trial, but it didn't measure collagen.

Bragato and colleagues (2025) gave 95 women a red LED mask two or three times a week for four weeks, against sham. A specialist wrinkle scale showed no difference between groups. Software measurements found shorter lines between the brows and around one eye, and satisfaction was higher in the light groups.

What the reviews say

Jagdeo and colleagues (2018) reviewed 31 randomised controlled trials of LEDs in dermatology. They were broadly positive but listed real problems: small sample sizes, missing blinding, some trials without a sham, and settings that varied from study to study.

Das and colleagues (2026) reviewed eight studies of at-home LED face devices with 365 participants. They found improvements in wrinkles, texture and elasticity with mild, short-lived side effects, and asked for larger, standardised trials with more people of darker skin types.

How to read this

There's a plausible mechanism and a small number of human studies that measured more collagen after light. There are more trials showing changes in how skin looks, some of them mixed. Nobody has run the large, long, independent trials that would show how much collagen changes, for whom, and for how long.

Who should check with a doctor first

If you have a skin condition, a changing mole or anything on your skin that worries you, see a GP first. Talk to your GP or pharmacist before using any device if you're pregnant, take medicine that makes your skin sensitive to light, have an eye condition, or have epilepsy triggered by flashing lights. Our pages on eye safety and side effects cover this in more detail.

What we won't say

We won't say our mask or panels increase collagen, firm skin or reverse ageing. Those claims would need trials of our own devices, and our specifications aren't confirmed. What we can say is what the research found, with its limits in plain view.