Home / Methodology

Methodology

Updated August 2026 · how this index is compiled, verified, labeled, and corrected

This index compiles published specifications for 188 red light therapy devices across 41 brands, labels every irradiance figure with the instrument that measured it, attributes independent measurements to their source by name, and ranks devices on four checkable criteria. Every clinical claim carries a PubMed ID that is verified programmatically before the site can publish. We do not operate a test lab, and we do not claim to: the value of this index is disciplined compilation and labeling of data that already exists, applied identically to every brand.

The four-factor rubric

Every ranking on this site is scored on the same four factors, and the same rubric is applied to every brand in the database, from the market leaders to single-panel storefronts.

1. Wavelength coverage and published density. Not the wavelength count on the box, but where the LED power actually sits. A "seven wavelength" panel can put 80 percent of its output into two bands; a brand that publishes its per-wavelength allocation makes that checkable, and a brand that publishes only a list is asking to be taken on faith. Published density earns credit; a long list by itself earns none.

2. Irradiance with a stated method. An irradiance figure is only meaningful as a triplet: the number, the instrument, and the distance. A solar power meter reads roughly double what a laboratory spectrometer records on the same panel, so a method-stated 70 mW/cm² can describe more real output than an unlabeled 200. Method-stated figures are compared like for like, method-unstated figures are discounted, and honestly low, spectrometer-verified numbers are treated as transparency rather than weakness.

3. Buyer terms. Trial length, restocking fees, whether the stand is included or a paid extra, and warranty length, all read from the policy fine print rather than the marketing banner. A 60-day free trial and a no-returns policy are different products at the same price.

4. Price per delivered hardware. What the verified price actually buys: LED count and chip architecture, coverage area, and included accessories.

No brand pays for a position. Affiliate relationships exist, are disclosed plainly, and never change a spec, a grade, or a ranking; category winners include brands with which we have no commercial relationship. See the affiliate disclosure.

The data protocol

The primary source for every record is the brand's own published product and policy pages. Each price in the database carries its own verification date, and a price whose date has gone stale is flagged for re-verification. Fields a brand does not publish are recorded as "not published" and rendered that way on device pages, never estimated: a blank cell is information about the brand's transparency, and inventing a value would destroy exactly the comparability this site exists to provide.

Independent measurements are layered on top of the manufacturer data, never blended into it. Where a third party has put a spectrometer on a panel, that figure appears with the source named inline: principally the published testing of Light Therapy Insiders, plus Outliyr and accredited lab reports where brands publish them.

Verification passes

In August 2026 the full database went through a two-stage audit. The first pass re-checked every metric on every device record against the manufacturer's current pages: wavelengths, density, irradiance and its method, LED counts, chip type, coverage, pricing, trial, restocking, stand, and warranty terms. That pass produced 777 field corrections and fills. The second pass then challenged the audit itself: every remaining "not published" marking was re-searched with AI-assisted lookups, and each candidate answer was verified against the underlying source before acceptance. Fifty-five markings were upgraded to sourced values; the rest were confirmed as genuinely unpublished.

Citation policy

Every PubMed ID that appears anywhere on this site is verified programmatically against PubMed before publication. The site generator keeps a register of verified PMIDs, each checked against the PubMed record's own title, and refuses to publish if an ID outside that register appears on any page: the build fails and nothing ships. If we cannot verify that a study exists and says what we cite it for, it is not cited.

Measurement-method policy

The single largest distortion in this market is instrumentation. A solar power meter responds to a broader band of light than an LED panel emits, so it reads roughly twice what a laboratory spectrometer records on the same panel at the same distance. Most headline marketing figures are solar meter numbers; most independent tests are spectrometer numbers; comparing one to the other is comparing instruments, not panels. This index therefore labels every irradiance figure with its instrument, treats unlabeled figures skeptically, and suggests halving a method-unstated headline number as a working assumption. The full explanation is at spectrometer vs solar meter, and every claim that can be checked against a spectrometer measurement is collected in the claimed vs measured dataset, downloadable as CSV.

Evidence grades

Condition pages grade the clinical evidence with four fixed labels, based on trial design, replication, and sample size rather than marketing volume:

Strong: replicated controlled human trials support the effect, as with skin, hair growth, wound healing, and knee joint pain.
Moderate: controlled human trials exist but are fewer, smaller, or less consistently replicated, as with muscle recovery and neuropathy.
Limited: early or small-sample evidence justifies interest but not confidence, as with sleep and psoriasis.
Insufficient: no credible controlled human support at consumer-device doses, as with weight loss, testosterone, and cognitive claims.

A grade describes the evidence for a condition, not any particular device; where trials used lasers or head-mounted devices rather than panels, the page says so.

Update cycle and corrections

Prices and specs are re-verified on a monthly pass, and the sitewide "updated" stamp reflects the last completed pass. Corrections from readers and brands ship within days: send the device name and a source link via the contact page, and the record's verification date updates with the fix. Brands can submit full spec sheets, including irradiance method and distance; inclusion is free and rankings cannot be bought.

Changelog

August 2026 - Initial publication: 188-device database across 41 brands, with method-labeled irradiance, per-wavelength density where published, and verified pricing.

August 2026 - Database-wide metric verification: every metric on every record checked against manufacturer pages, producing 777 field corrections and fills, followed by a source-verified challenge pass that upgraded 55 "not published" fields to sourced values.

August 2026 - Citation register added: every PMID on the site is programmatically verified against PubMed at build time, and an unverified ID stops publication.

August 2026 - Independent lab attribution: third-party measured irradiance figures are attributed to their source by name wherever they appear.

August 2026 - Site-wide accessibility and performance pass: breadcrumbs, mobile-scrollable tables, image fallbacks, long-lived caching, self-hosted fonts, and WebP variants.

August 2026 - Filterable all-panels page added: the full database in a single list with brand, wavelength, and price filters.

August 2026 - Claimed vs measured study published: every irradiance claim in the database that can be checked against a spectrometer measurement, with per-panel ratios and a downloadable CSV.