Red Light Therapy: The Skin Data Is Real, the Longevity Claims Aren't

The full story

In June 2025 the FDA cleared an over-the-counter LED light therapy mask with exactly one indication for use: the treatment of full-face wrinkles. That is the entire cleared claim, on a Class II device. Not mitochondrial support. Not cellular energy. Not healthy ageing. Wrinkles, on a face.

The distance between that sentence and the copy on a typical red light panel is what this piece is about. The frustrating part is that the underlying science is not junk. Photobiomodulation has randomised trials, a plausible molecular target and a dose-response curve more interesting than anything in the marketing. It simply does not go where the marketing says it goes.

The mechanism is real, and it is narrow

The proposed photoacceptor is cytochrome c oxidase, and there is a human demonstration. A 2017 study in the Journal of Cerebral Blood Flow and Metabolism applied a 1064 nm laser to the foreheads of 11 healthy participants and measured a significant increase in oxidised cytochrome c oxidase, more than 0.08 micromolar, along with rises in oxygenated and total haemoglobin, versus placebo sessions.

Read that carefully. Eleven people, a spectroscopic measurement of a redox state, a genuine effect. It is a biochemical change measured in a laboratory, and it is the load-bearing evidence beneath an enormous quantity of "supports your mitochondria" copy. Nothing in it concerns how anyone felt, functioned or aged.

Where the randomised evidence holds up

Skin is the strongest case. A 2014 controlled trial in Photomedicine and Laser Surgery treated 136 volunteers, 113 of them randomised across four treatment groups against 23 controls, twice weekly for 30 sessions, using either 611 to 650 nm red light or 570 to 850 nm polychromatic light normalised to about 9 J/cm2. Treated subjects had significantly better profilometrically assessed skin roughness, ultrasonographically measured intradermal collagen density and blinded photographic evaluation. Notably, the broadband spectrum offered no advantage over red alone.

A 2023 split-face randomised trial in Photobiomodulation, Photomedicine and Laser Surgery sharpened the picture. In 137 women aged 40 to 65, ten sessions over four weeks at 3.8 J/cm2 per side reduced periocular wrinkle volume by 31.6 percent with 660 nm red and 29.9 percent with 590 nm amber. Two things follow. The effect is narrow: neither wavelength improved skin hydration or viscoelasticity. And at matched dose, red and amber were near-identical, which is awkward for an industry that sells wavelength as the secret ingredient.

Pain has decent evidence too. A 2009 meta-analysis in the Lancet pooled 16 randomised trials of low-level laser therapy in neck pain across 820 patients. The relative risk of pain improvement versus placebo was 1.69 in acute neck pain and 4.05 in chronic neck pain, with pain intensity falling 19.86 mm on a 100 mm visual analogue scale and relief persisting between 1 and 22 weeks after treatment stopped. Side effects were mild and no different from placebo. A 2019 meta-analysis in BMJ Open covering 22 placebo-controlled trials and 1,063 patients with knee osteoarthritis found pain down 14.23 mm overall, but 18.71 mm when only the doses recommended by the World Association for Laser Therapy were pooled, peaking at 31.87 mm beyond placebo at two to four weeks of follow-up. Those same authors note that low-level laser therapy is not recommended in major knee osteoarthritis guidelines.

The trials that failed, including one where it hurt more

A 2023 meta-analysis in PLoS One pooled 18 randomised trials and 793 participants in rheumatoid arthritis and concluded, on low-quality evidence, that there may be no difference between infrared laser and sham for pain, morning stiffness, grip strength, functional capacity, inflammation, range of motion, disease activity or adverse events. Evidence for red laser and laser acupuncture was very uncertain.

A 2018 randomised trial in Clinical Interventions in Aging assigned 68 patients with lumbar disc degenerative changes to high-intensity laser at 1064 nm and 60 J/cm2, low-level laser at 785 nm and 8 J/cm2, or matching sham. All four groups improved on pain and disability scores. None of the active arms beat placebo at three weeks or at one and three months. That is a textbook demonstration of how much apparent benefit in this field is regression, attention and the passage of time.

And a 2017 triple-blind randomised trial in Lasers in Surgery and Medicine gave 54 postpartum women three sessions of real or simulated low-level laser therapy after mediolateral episiotomy. Healing scores did not differ, and pain scores were statistically higher in the laser group at the first assessment and again after the third session. Dose does not only fail to help. It can point the wrong way.

Proponents have an answer for negative trials, and it is a legitimate one. A 2011 review in Dose-Response attributes a substantial share of null low-level-light results to inappropriate dosimetric parameters. That same review documents the biphasic dose-response that makes this whole field awkward: ATP and mitochondrial membrane potential rise then fall as dose increases, mitochondrial reactive oxygen species show a triphasic pattern with two peaks, and transcranial photobiomodulation in mice peaks and then declines as you increase either the number of treatments or the energy density per treatment. More light is not better. But notice the escape hatch this creates. Any failure can be attributed to the wrong dose, which is the structure of an unfalsifiable claim unless the right dose is specified in advance.

What the evidence does not show

Nothing in this literature touches lifespan, healthspan or any hard clinical outcome. There is no trial here showing that shining red light on your torso changes how you age, because those trials have not been run. The chain being sold to you, in which light hits cytochrome c oxidase, mitochondria make more ATP, cells therefore age more slowly and you therefore age more slowly, snaps at the second link. A measurable rise in oxidised cytochrome c oxidase across eleven foreheads is not a longevity result, and the only regulatory clearance in evidence here is cosmetic and facial.

It also does not show that consumer panels and masks are interchangeable with the devices used in the trials, because most consumer devices decline to publish the two numbers that would let you check.

The only specs that carry information

A 2012 review in Annals of Biomedical Engineering states it flatly: wavelength, light source, irradiance and delivered energy density are what determine whether a treatment works. In practice that means two figures.

• Irradiance, in mW/cm2, at a stated distance from the panel.
• Delivered dose, in J/cm2, which is what irradiance and session time combine to produce.

Total wattage, diode count and the phrase "medical grade" are not specifications. And the doses that produced effects in trials are modest and repeated rather than heroic: about 9 J/cm2 twice weekly across 30 sessions in the skin trial, 3.8 J/cm2 per side across ten sessions in four weeks in the split-face trial, and for joints, 4 to 8 J at 785 to 860 nm or 1 to 3 J at 904 nm per treatment spot. The 60 J/cm2 arm in the lumbar trial did nothing at all. A reasonable person works out the session time that lands them in the studied range at their device's stated irradiance and distance, then stops, because on a biphasic curve doubling the dose is not a bonus. It is a different experiment.

Red light therapy occupies the unusual position of being undersold and oversold at once. The wrinkle and neck-pain data are better than a sceptic expects. The longevity story is thinner than an enthusiast will admit. And the deciding variable in nearly every case is a dose figure that most sellers decline to print. If a device tells you its irradiance and its delivered dose, you can compare it against a trial and decide for yourself. If it tells you about your mitochondria, it is telling you about its marketing department.