A red light therapy protocol is four variables, not a device. Wavelength, dose, frequency and target tissue. Fix those four and session length falls out of arithmetic. Copy someone else's session length without them and you have copied nothing, because their irradiance, their distance and their target tissue were not yours.
The harder half: nobody has established the correct four numbers for any consumer use. Every range you will see quoted describes what researchers happened to use, or where positive results happened to cluster. So this page will not hand you a recipe. It teaches you to derive a number you can defend and repeat, and says plainly how wide the error bars are.
General information, not medical advice.
The four variables
1. Wavelength is a band, not a line
The literature concentrates on red at roughly 630 to 670nm and near infrared at roughly 780 to 940nm, but those edges are not agreed. Other sources put red at 620 to 700nm and near infrared anywhere from 700nm to 1440nm; muscle trials span 655 to 950nm, hair trials 600 to 1100nm. Read a stated wavelength as what one study used, not as a category with fixed meaning. An LED labelled with one wavelength also emits across a band: for a 20nm bandwidth, modelling puts only about 4.7 percent of the quoted dose at the stated central wavelength. The head to head trials are unimpressive too. 660nm against 808nm on muscle fatigue and 810nm against 940nm after dental surgery both found no difference between wavelengths.
2. Dose is the number everyone quotes and nobody has validated
Dose, or fluence, is joules per square centimeter, and the published ranges disagree. The most cited review says 1 to 20 J/cm2. A second review from an overlapping author team gives 1 to 10 J/cm2 in one place and hedges perhaps 1 to 100 J/cm2 in another, in the same paper. And a 2025 audit of 27 visible LED dermatology studies found real trials using 0.1 to 126 J/cm2, median 40.5, even between studies treating the same condition. A device advertising an optimal dose is quoting a number the field has never agreed on.
Units make it worse. Consensus dosing for 904nm pulsed lasers is 1 to 4 joules per treatment point with a 100 mW/cm2 maximum at several named sites, and it is written for clinician applied class 3B lasers on named joints and tendons, not for LED panels at 660 or 850nm. Muscle trial doses are total joules per muscle group, roughly 20 to 60 J small and 60 to 300 J large. Those ranges describe where positive results clustered in a literature the review itself rates very low to moderate quality, not a window that works: a blinded 2025 trial of 60 joules, 300 joules and an escalating dose found all three no different from sham, and an earlier trial found 30 joules null between positive 10 J and 50 J arms. No conversion exists between joules per point, joules per muscle group and J/cm2 at the skin.
3. Frequency has almost no evidence behind it
No evidence establishes a dose response for treatment frequency in humans. The one comparison in facial skin found three sessions a week no better than two. The only animal data is a non significant trend toward worse outcomes in mice given 14 consecutive daily treatments, which argues against assuming more is better rather than showing harm. Chronic pain trials most commonly ran 2 to 3 sessions per week and home hair trials 3 to 4 times weekly, conventions nobody compared. The schedule page takes that apart.
4. Target tissue is where the numbers stop being honest
Every joule figure in the consumer literature is a surface number, and what arrives deeper is a small, largely unmeasured fraction. Published penetration figures run from about 4 to 5mm to about 40mm and do not measure the same thing: 4 to 5mm is a model's one percent energy contour, 14 to 26mm is lateral spread through living skin, 40mm is where 808nm stayed detectable in a cadaver head with no circulating blood. Those modelling authors say plainly that a therapeutic depth cannot be identified from their data. Nor does depth rise neatly with wavelength: in that cadaver work 808nm beat both 660nm and 940nm.
Worked example: solving for the fourth number
The arithmetic is a definition, not a finding. Fluence in J/cm2 equals irradiance in W/cm2 multiplied by time in seconds, so fix a target dose and a real irradiance at your real distance and time is the only unknown left.
- Target dose: 10 J/cm2, from the 1 to 20 J/cm2 range in the most cited review. Record whose convention it is, because three other published ranges give a different answer.
- Irradiance at your distance: 50 mW/cm2, which is 0.05 W/cm2.
- Time: 10 divided by 0.05 is 200 seconds, which is 3 minutes 20 seconds, not the 3 minutes you would round it to.
Now change one input. If what actually reaches your skin at that distance is 15 mW/cm2, the same target takes 667 seconds, about 11 minutes. Same dose, more than triple the session. That is why a session length quoted on its own is worthless advice.
Irradiance is the weak link, and it is the number you have least right to trust. Manufacturers quote a figure measured at the panel face, no federal performance standard applies to non laser LED panels, and nobody has independently measured a full-size consumer panel for delivered irradiance at all. The only bench measurements published anywhere are of five hand held LED units, and whether a large panel behaves the same way is unmeasured.
Two things also stop you shortcutting the arithmetic: reciprocity failure, and a dose ceiling documented in animals and never located in people. Both mean you may not be able to move closer and halve the time. The session length page works both through in detail.
BEYOND SUPPLEMENTS No stack adds muscle you did not earn. EMS might. Twenty minutes, twice a week, every major muscle group under load. Here is what the studies actually measured. READ THE RESEARCH →The only guideline grade protocol that exists anywhere
One place in medicine has written down red light parameters and told clinicians to follow them. The 2019 MASCC and ISOO mucositis guidelines recommend intra-oral photobiomodulation by low level laser to prevent oral mucositis in cancer patients, at Level of Evidence I in stem cell transplant and in head and neck radiochemotherapy, and at Level of Evidence II in head and neck radiotherapy without chemotherapy. That is the whole list.
The specificity is the point. One recommended stem cell transplant protocol is 632.8nm at 31.25 mW/cm2, 40 seconds per spot, 1.0 J/cm2, an 0.8 cm2 spot, 18 sites, and that 1.0 is the guideline's own stated figure even though 31.25 mW/cm2 for 40 seconds multiplies out to 1.25, an inconsistency worth noticing in the only guideline that exists. Another, equally recommended, is 650nm at 1000 mW/cm2 for 2 seconds per spot, 2.0 J/cm2, a 0.04 cm2 spot, 54 to 70 sites. The radiochemotherapy protocols run 660nm at 417 mW/cm2, 4.2 J/cm2 over 0.24 cm2, or 660nm at 625 mW/cm2, 6.2 J/cm2 over 0.04 cm2. Clinicians are told to follow the parameters of the protocol they choose rather than treat the numbers as interchangeable, and 40 seconds at 31.25 mW/cm2 sitting beside 2 seconds at 1000 mW/cm2 shows why.
None of it has been shown to transfer to a panel. These are clinician delivered laser spots of 0.04 to 1 cm2 at up to 1000 mW/cm2, roughly 20 to 200 times a consumer panel, aimed inside a mouth, for a condition no home device is sold for. Whether any guideline protocol transfers to a panel, mask or bed has never been tested, and the formats differ by two orders of magnitude in irradiance and three in treated area. The numbers are contested inside their own field too: a position paper proposes a working envelope of 5 to 150 mW/cm2 with 1 to 6 J/cm2 called safe and effective, and three of the five MASCC protocols sit far outside it. Worth borrowing from that paper is its thirteen item reporting checklist: wavelength, power, beam area, irradiance, radiant energy, treatment duration, fluence, continuous wave or pulsed mode, pulse structure, physical relationship to the target tissue, timing relative to other therapy, schedule and frequency, and anatomical sites treated. Almost no consumer device publishes enough to complete half of it.
Protocol by goal
Parameters, not verdicts. For evidence strength by outcome, see the red light therapy guide.
| Goal | Target tissue | Wavelengths in the trials | Format used in the trials | Where the numbers break |
|---|---|---|---|---|
| Facial skin | Epidermis and upper dermis | 630nm, 660nm, sometimes 850nm added | Mask or panel at a fixed distance | No dose ranging trial has ever been run in human facial skin. |
| Hair | Follicle in the dermis | 600 to 1100nm across all trials including in-office lasers; cleared home devices span 620 to 678nm, most at 655 or 660nm | Comb, cap, helmet or band worn in contact | Sessions in sham controlled trials ran 90 seconds to 36 minutes, and no trial compares device designs. |
| Muscle recovery | Muscle belly, under skin and fat | 655 to 950nm, maximum 200 mW per diode | Laser or LED clusters held against one muscle | Doses are total joules per muscle group, which do not convert to J/cm2. The only whole body panel review found none of its five studies, covering 105 participants, reported any benefit to performance or fatigue biomarkers, though two reported better sleep quality. |
| Joint and back pain | Joint capsule, tendon, deeper structures | 660 to 905nm | Contact probe on named points, by a clinician | Trials ran 1 to 20 sessions at 4 to 130 J/cm2. Transfer to a panel at a distance is untested. |
How to derive your own numbers
- Name the target tissue and its depth. Nothing establishes that a dose working at the surface relates to one working deeper.
- Pick a wavelength band from trials in that tissue. Do not pay for extra wavelengths on the theory that more is better: the only direct comparison of combined versus single wavelengths is a rat study that found no difference between the two treatments, and no human trial has tested it.
- Get an irradiance figure at your actual distance, not off the box. If you cannot, hold the distance constant so the error at least stays constant.
- Choose a target dose and record which published range you took it from. This is the step that stops you drifting.
- Solve for time, and do not shorten the session by moving closer.
- Hold all four fixed and change one at a time. The results timeline page covers how long to hold before judging, and when in the day to run it covers the one variable that genuinely changes around training. If you have not bought hardware, that fixed distance problem is the strongest argument for one format over another, which our device comparison covers.
Before you copy anyone's numbers
Photosensitizing medication. Most drug induced photosensitivity runs through ultraviolet A at 315 to 400nm, which a red panel does not emit. But the same review lists drugs that react to visible light across 400 to 740nm, a band that contains red, individual action spectra vary between people, almost no photosensitising drug has been tested against red or near infrared light directly, and many masks add blue LEDs around 415nm. The clearest documented cases are not in doubt: porfimer sodium is activated at 630 plus or minus 3nm, inside the red band, and verteporfin at 689 plus or minus 3nm, with a label naming light emitting medical devices among sources to avoid for five days after infusion. So the blanket warning is broader than the UVA mechanism justifies, and it is not something to overrule yourself. Ask your prescriber.
Light sensitive conditions. The photodermatoses are mostly ultraviolet driven, but solar urticaria and cutaneous lupus show visible light reactivity, and protoporphyrin keeps a weak absorption band at 629nm, so porphyria and recent photodynamic therapy are not conditions to wave through.
Pregnancy. No study of consumer LED panels or whole body red light exposure in pregnancy exists. The manufacturer contraindication is a default rather than a finding, which is a reason for caution, not reassurance.
Eye protection. Wrong when stated flatly in either direction. Visible red triggers a normal aversion response, and red light has been delivered deliberately into human eyes in a clinician operated retinal trial whose 13 month interim analysis reported a favorable safety profile, though six of its authors work for the device maker and that same trial delivered 850nm near infrared into the same eyes. None of it describes a consumer panel at home. Near infrared triggers no aversion response, which is why ICNIRP writes a separate retinal limit for sources giving no strong visual stimulus, and its corneal limit for the 780 to 3000nm band is 10 mW/cm2 for exposures of 1000 seconds or longer. That governs what arrives at the eye, not what a panel emits, so compliance cannot be computed from a marketing figure. Use eye protection with near infrared, and do not stare into any of it.
General information, not medical advice. If you are pregnant, take photosensitizing medication, or have a light sensitive condition, talk to a doctor before starting.
FAQ
What is the standard red light therapy protocol?
There isn't one. The only guideline grade protocols anywhere are the MASCC and ISOO intra-oral laser protocols for preventing oral mucositis in cancer patients, delivered by a clinician to spots of 0.04 to 1 cm2 at up to 1000 mW/cm2. No clinical practice guideline covers red or near infrared light for skin, hair, performance, sleep or general wellness.
Why does the same protocol give two different session lengths?
Because session length is not one of the four variables, it is what falls out of them. Fluence in J/cm2 equals irradiance in W/cm2 multiplied by time in seconds, so once the target dose is fixed, the time is decided by the irradiance actually arriving at your skin, and that changes with your device and your distance. Two people running the identical four variables can end up minutes apart. The session length page takes the distance and dose ceiling problems apart.
Is 10 J/cm2 the optimal dose?
No number has been shown to be optimal. Published ranges variously say 1 to 20 J/cm2, 1 to 10, or perhaps 1 to 100, while an audit of 27 visible LED dermatology studies found trials using 0.1 to 126 J/cm2. Positive results have been reported at the top of that spread as well as far below it, and trials landing inside the published muscle dose windows have returned nothing.
Can I use one protocol for skin and for muscle?
Not with any confidence, because the two sit at different depths and every joule figure published is a surface number. Facial skin trials cluster at 630nm and 660nm with the target a millimeter or two down. Muscle trials run 655 to 950nm with the target under skin and fat, and dose in total joules per muscle group, a unit that does not convert to J/cm2 at the skin. Published penetration estimates run from about 4 to 5mm to about 40mm and do not measure the same thing, and the modelling authors say plainly that a therapeutic depth cannot be identified from their data.
Do I need eye protection?
It depends on the output. Visible red produces a normal aversion response and has been delivered directly into human eyes in a clinician operated retinal trial whose 13 month interim analysis reported a favorable safety profile, though six of that trial's authors work for the device maker and it delivered 850nm near infrared into the same eyes, so it says nothing about a consumer panel at home. Near infrared produces no aversion response, which is why international guidance writes a separate retinal limit for it. No threshold has been established at which protection becomes necessary for a consumer device, so the cautious position is goggles with near infrared and no staring into anything.
Frequently asked questions
What is the standard red light therapy protocol?
There isn't one. The only guideline grade protocols anywhere are the MASCC and ISOO intra-oral laser protocols for preventing oral mucositis in cancer patients, delivered by a clinician to spots of 0.04 to 1 cm2 at up to 1000 mW/cm2. No clinical practice guideline covers red or near infrared light for skin, hair, performance, sleep or general wellness.
Why does the same protocol give two different session lengths?
Because session length is not one of the four variables, it is what falls out of them. Fluence in J/cm2 equals irradiance in W/cm2 multiplied by time in seconds, so once the target dose is fixed, the time is decided by the irradiance actually arriving at your skin, and that changes with your device and your distance. Two people running the identical four variables can end up minutes apart.
Is 10 J/cm2 the optimal dose?
No number has been shown to be optimal. Published ranges variously say 1 to 20 J/cm2, 1 to 10, or perhaps 1 to 100, while an audit of 27 visible LED dermatology studies found trials using 0.1 to 126 J/cm2. Positive results have been reported at the top of that spread as well as far below it, and trials landing inside the published muscle dose windows have returned nothing.
Can I use one protocol for skin and for muscle?
Not with any confidence, because the two sit at different depths and every joule figure published is a surface number. Facial skin trials cluster at 630nm and 660nm with the target a millimeter or two down. Muscle trials run 655 to 950nm with the target under skin and fat, and dose in total joules per muscle group, a unit that does not convert to J/cm2 at the skin. Published penetration estimates run from about 4 to 5mm to about 40mm and do not measure the same thing, and the modelling authors say plainly that a therapeutic depth cannot be identified from their data.
Do I need eye protection?
It depends on the output. Visible red produces a normal aversion response and has been delivered directly into human eyes in a clinician operated retinal trial whose 13 month interim analysis reported a favorable safety profile, though six of that trial's authors work for the device maker and it delivered 850nm near infrared into the same eyes, so it says nothing about a consumer panel at home. Near infrared produces no aversion response, which is why international guidance writes a separate retinal limit for it. No threshold has been established at which protection becomes necessary for a consumer device, so the cautious position is goggles with near infrared and no staring into anything.