Red Light Therapy Results Timeline: Four Clocks

There is no single timeline, and the one you were given was probably borrowed from the wrong tissue. Skin, hair, muscle recovery and pain run on four separate clocks. The fastest is measured in hours after one session, the slowest in half a year. Applying the hair number to your face, or the skin number to your quads, is how people quit three weeks early or grind on for a year after the evidence stopped saying anything.

And one thing to know before the numbers below: almost every timeline here is a measurement schedule, not a record of when anyone noticed something. Researchers picked week 4, week 8 or week 26 because that is when the camera was booked. No sham controlled trial in the hair or skin literature reports the earliest week a visible difference appeared.

The four clocks, and why none of them is week by week

Target	What a trial course looked like	Earliest separation from sham on record	What nobody measured
Skin	5 sessions a week for 12 weeks, or twice weekly for about 15 weeks in an older trial with no sham arm, run in gas discharge lamp cabins rather than LEDs and funded by the cabin maker	Week 8, in the one home mask trial assessing at weeks 4, 8, 12 and 16	What happens after you stop, outside one uncontrolled 20 person study
Hair	16 to 26 weeks, mean 21.3, sessions of 90 seconds to 36 minutes	Not established. Trials report the endpoint, not the week	Any follow up after discontinuation
Muscle recovery	A single application before or after one session, and the longest tested course was ten sessions over five weeks	72 and 96 hours for reported soreness, 24 and 48 hours for strength, after a single damaging session	Anything past six weeks of training, and the one trial that looked found no added benefit
Pain	1 to 20 sessions across 3 days to 12 weeks	End of course, and contested: one meta-analysis of 22 knee trials still shows an effect 2 to 12 weeks later, an earlier one of nine found nothing	Almost nothing past 8 weeks, and the one 6 month follow up is manufacturer funded and entirely self reported

Skin: nothing a clinician scores at four weeks, an answer around eight

The most useful trial is a sham controlled, double blind study of a home mask in 60 Korean adults, using 630nm red plus 850nm infrared at a maximum of 10 mW/cm2 for 9 minutes, five times a week for 12 weeks, assessed at weeks 4, 8, 12 and 16. It was funded by the mask's maker, Y&J Bio, and the sham device was not dark: it emitted 630nm at one tenth intensity, so the comparison is full dose against a tenth of a dose, not against no light. At week 4 the groups did not separate. Significant differences in crow's feet grading appeared at weeks 8, 12 and 16, ending with 86.2 percent of the active group improving by at least one point versus 16.7 percent of sham.

A second sham controlled trial, 95 women using a 660nm mask at 6.4 mW/cm2 and 8.05 J/cm2 for 21 minutes over 4 weeks, found no significant difference from sham on the clinician rated Wrinkle Assessment Scale. Not wholly negative: software measurement found reductions in wrinkle length, and satisfaction was higher in both active groups. But the validated scale showed nothing at four weeks.

You will also meet a widely quoted 31.6 percent periocular wrinkle volume reduction at four weeks, from a split face trial in 137 women at 3.8 J/cm2. Both sides of every face received active light, red 660nm against amber 590nm. So that figure is a change from each woman's own baseline, not a difference against a control, and the comparison the design actually tested came out roughly even.

The honest gap: there is no washout data. Nobody has run a controlled study of what happens to skin after you stop. The only persistence measurements in existence were taken 14 and 28 days after stopping, in an open label study of 20 volunteers with no control group and the device maker's chief executive among the authors. That is not evidence gains hold, and not evidence they fade. It is an absence.

Hair: the slowest clock, and the one where the instrument may be the only thing that notices

Every sham controlled trial supporting these devices ran 16 to 26 weeks. A review of seven double blind randomized trials in 607 participants put the mean at 21.3 weeks, with sessions from 90 seconds to 36 minutes.

The part that never makes it into marketing: no trial reports the week at which hair density first separates from sham. The 26 week figure is when the phototrichogram was taken, not when the effect started. Anyone telling you month three is when it kicks in is describing a study design that does not exist.

The endpoint: at 26 weeks, treated arms in the multicenter laser comb trials gained 18.4 to 25.7 terminal hairs per square centimeter. Sham arms gained too, 1.6 to 9.4. So the gain attributable to the device is the difference, roughly 17 to 18 hairs, close to Cochrane's pooled mean differences of 17.40 and 17.60.

Then the finding that should shape expectations more than any of the above. In that same Cochrane analysis, participants did not rate the laser comb as more effective than the sham device, at relative risks of 1.54 and 1.18, both non significant, across two studies totalling 141 participants. The counts moved. The women holding the device could not tell. And Cochrane graded the participant rated evidence moderate quality and the hair count evidence low quality, so the half of that result you can check for yourself is the better evidenced half, and it was the null one. Against that sits a retrospective dataset of 1383 users in which 20 to 23 percent showed no effect at all, but it is uncontrolled, self reported and co-authored with the manufacturer, so on the same subjective question it is weaker evidence than Cochrane, not a rebuttal. Expect a change an instrument can measure and you may not see.

The honest gaps: no trial follows anyone after they stop, so nothing is known about how fast counts return to baseline. The European S3 guideline explicitly cannot recommend for or against use beyond six months. The longest published follow up, 48 weeks, was open label with no control arm.

Muscle recovery: hours, not weeks, and it probably does not compound

This clock is different in kind. Effects, where they exist, land in the hours after one hard session rather than building over a course. A 2025 meta-analysis found moderate reductions in reported pain at 72 and 96 hours after damaging exercise, at standardized mean differences of 0.55 and 0.56. That pooled only 4 of the review's 14 studies, about 284 people, with four of the included studies finding nothing and the pain self reported in trials where participants could see whether the light was on. The same review put strength recovery at 24 and 48 hours at 0.97 and 0.99, but a larger pool of 34 trials put that construct at 0.24, and the lower bound of the 48 hour interval reaches 0.05, so the data are compatible with almost nothing. Take the smaller number. So the timepoints where anything was found are 72 and 96 hours for reported soreness and 24 and 48 hours for strength, and the strength figures are the ones with the interval reaching almost to zero.

The multi week version keeps coming back empty. Six weeks of sprint and squat training with light added produced no benefit over placebo in trained men, and nothing in the literature shows an effect that builds across a block. There is a positive side to weigh that against: the athlete subgroup of a 34 trial meta-analysis did show a benefit, alongside pooled findings of about four extra repetitions in an acute resistance set and reduced soreness, all of it from contact clusters held against a muscle and all of it from the same low quality pool. The strength of the recovery evidence overall is ranked in what the evidence supports.

If a run of hard sessions produced nothing, the delivery format is a likelier explanation than the calendar, which the timing page covers.

Pain: a course of weeks, then a measurement blackout

Across 14 randomized trials of photobiomodulation for chronic pain, courses ran from one session to 20, across 3 days to 12 weeks, most commonly 2 to 3 a week. Follow up usually ended within 8 weeks of the final session. Past that, almost nothing is measured. One exception exists and it comes with its own asterisk: a light bed trial in 42 fibromyalgia patients reported pain and quality of life still improved against placebo six months after the last session, but every outcome was a self reported questionnaire and two of the four authors worked for the company that makes the bed.

Within a course, knee osteoarthritis is the best documented case and it is contested. A meta-analysis of 22 placebo controlled trials in 1063 people found pain reduced by 14.23mm on a 100mm scale at end of therapy, rising to 23.23mm at 2 to 12 weeks afterwards, an odd shape in which the effect grew after the sessions stopped. An earlier meta-analysis of nine trials concluded the opposite, at a standardized mean difference of minus 0.28 with an interval crossing zero. Read them together. The positive one also depended on a clinician pressing a contact probe onto named points at 4 to 8 joules per spot at 785 to 860nm, which is not what a panel does, and the authors of that review open by noting the treatment is not recommended in the major knee osteoarthritis guidelines. In chronic low back pain, seven trials showed reduced pain scores with no improvement in disability: what people reported moved, what they could do did not.

When it is reasonable to stop

Decide this before you start, because what keeps people paying is a treated group number from a trial where the sham group also improved.

Skin: stop at 12 weeks. The full treatment course in the only trial that pins separation from sham to a week, where separation was already there at week 8 and the week 16 reading was a follow up taken a month after treatment had ended. Twelve weeks of a schedule you actually kept, with nothing you or anyone else can see, is a complete answer.
Hair: stop at 26 weeks. The outer edge of every sham controlled trial ever run on these devices, and beyond it the European guideline will not recommend for or against. Judge it on photographs taken under the same light, not on the mirror, since the blinded participants in Cochrane could not tell.
Recovery: judge it session by session. Any effect, where one exists, shows up in the days after a hard session, at 72 and 96 hours for soreness and 24 and 48 hours for strength. Nothing shows an effect that builds, so there is no course length to complete: if you cannot detect a difference in soreness or next day output, more weeks will not manufacture one.
Pain: stop at 12 weeks or 20 sessions. The outer limits of the trial courses, and most were considerably shorter.

None of these are pass marks. They are the point past which the evidence has nothing further to say, so continuing becomes a decision about habit rather than data.

Safety, on any of these clocks

Most photosensitising drugs act through ultraviolet A at 315 to 400nm, but the same source puts some drug reactions in the visible range out to 740nm, which overlaps the red end of a panel. Action spectra also vary by individual, and many masks add a blue channel around 415nm. So this is a pharmacist conversation, not a green light. Two drugs are unambiguous: porfimer sodium is activated at 630 plus or minus 3nm, with a label requiring bright light avoidance for at least 30 days and some patients still photosensitive at 90 days or beyond; verteporfin at 689 plus or minus 3nm, its label naming light emitting medical devices during a five day window. Light-sensitive conditions such as solar urticaria and cutaneous lupus show documented visible light reactivity, and no study has isolated red wavelengths in them. In pregnancy nobody has studied consumer panels, so the manufacturer contraindication is a default, not a finding.

Wear eye protection. Near infrared is invisible, so the blink and squint reflex that protects you from a bright red panel never fires, which is why radiation protection guidance sets a separate retinal limit for sources with no visual stimulus. A different limit, this one for the cornea and lens across 780 to 3000nm, is 10 mW/cm2 for exposures of 1000 seconds or longer, and that describes light arriving at the eye, not what a panel is rated to emit. The one published mask retinopathy case involved a 460 to 470nm blue channel used with the eyes open and no protection.

General information, not medical advice.

If your clock is running and nothing is happening, the next question is whether the dose was ever right: start with building a protocol from four variables and how long one session should run. Weekly structure matters more than any single session, and what the evidence supports is worth reading before committing six months to a clock this slow.

FAQ

How long until red light therapy works?

It depends on the tissue. For skin, one sham controlled trial showed no difference at week 4 and a significant one at week 8. For hair, every controlled trial ran 16 to 26 weeks and none reports when the difference first appeared. For recovery, the timepoints where anything was found are 72 and 96 hours after a session for soreness and 24 and 48 hours for strength. For pain, courses ran 3 days to 12 weeks.

Should I expect to see anything in the first two weeks?

No, and no trial supports it. The earliest any sham controlled skin trial has shown a difference is week 8, and two separate trials found nothing at week 4 on the validated clinician scales they used, though one of them did pick up shorter wrinkles on software measurement.

Do results fade after I stop?

Nobody knows. For skin, the only persistence data are measurements taken 14 and 28 days after stopping in an uncontrolled 20 person study with manufacturer involvement in authorship. For hair, no trial has followed participants after discontinuation, and no maintenance schedule exists for any indication.

Is shedding in the first month a sign it is working for hair?

There is no good evidence for it. The claim rests on a single 2003 report in a non-indexed journal, and telogen effluvium does not appear among the adverse events in the major sham controlled trials. Treat unexplained shedding as a reason to see a dermatologist, not a milestone.

Why would a trial measure a change I cannot see?

Because that is what happened in the Cochrane review of female pattern hair loss. Pooled hair counts favored the laser comb at mean differences of 17.40 and 17.60, while blinded participants rated it no better than sham, at relative risks of 1.54 and 1.18. Cochrane graded the participant ratings moderate quality and the hair counts low quality, so the half of that result you can see for yourself is the better evidenced half, and it was null.

At what point is it reasonable to give up?

Twelve weeks for skin, 26 weeks for hair, 12 weeks or 20 sessions for pain. Each of those is the outer edge of what has been trialled, so past that there is no evidence left to appeal to. Recovery has no course to complete, because nothing shows an effect that builds, so judge it session by session instead.

Frequently asked questions