Methodology

Last reviewed July 2026.

LED Mask Score checks two kinds of marketing against two kinds of evidence. The ranking checks the numbers masks are sold on — LED counts, claimed mW/cm² — against independently measured irradiance, cited to the spectrometer review behind every figure. The clearance record checks the two words every mask is sold on — FDA cleared — against the FDA's own public paperwork: the 510(k) database (every clearance's K-number, holder, device name, and decision date) joined with the registration & listing index (which marketed product names ride which K-number). Every K-number we cite links the FDA's record page so you can read it yourself. Where the two databases disagree or lag each other, we check both before saying anything — a listing without a K-number field is not proof there's no clearance, and we treat it that way.

The measured ranking — cited, never ours

We run no tests and publish no measurement without a citation. The ranking's figures come from Light Therapy Insiders' spectrometer tests (Alex Fergus) — one tester, one instrument, measured at the mask surface, with the delivered dose (fluence, J/cm²) read over the mask's own default session. The specific review behind each number is linked on its row with its publication date. Three rules keep the column honest. One ruler: a mask enters the ranking only with a measurement under this protocol — numbers metered by different testers at different distances aren't comparable, so masks with only foreign-protocol measurements wait in the unranked section, however interesting their numbers. Disclosed arithmetic: where a review prints dose per session rather than irradiance, the mW/cm² we show is that dose divided over its measured window — arithmetic on cited figures, never a new measurement, and flagged as derived on the row. The label never ranks: LED counts and claimed outputs are quoted exactly (checked 2026-07-16, source per row) and rendered muted beside the measurement. That contrast is the site's point — the 648-LED mask ranks third by dose, and a mask advertising 70 mW/cm² metered 5–11 in the cited review. Delivered dose still varies with fit and wearing distance; the measurement describes the mask under the cited protocol, not your face.

The verdict tiers

Two failure modes the tiers deliberately surface: a claim whose matched record covers a different indication than the marketing sells (a device cleared for acne marketed for wrinkles), and marketplace listings that ride a real K-number while never stating the model you'd receive. Both read as "FDA cleared" on the page; neither is the same fact as a brand-held clearance for the thing in the cart. And a phrase we never rank at all: "FDA registered" means a facility is on a list — no device was reviewed. A mask marketed on registration alone would sit below the line with the unmatched claims.

What gets ranked

A row is ranked only when three things are on file: a measured irradiance under the cited protocol, an FDA claim matched to a real K-number (first two tiers), and a verified first-party USD price with source and date. No cited measurement, no rank; no matched K, no rank; no first-party price, no rank — those rules keep several genuinely cleared masks out of the ranking today, and we'd rather have the gap than the guess. The ranking sorts by measured irradiance by default and never by anything we can't cite. The clearance record keeps its own table of every claim we've matched, sorted by price. The full datasets — every source URL, claim quote, review link, and price date — are public at data/rankings.js and data/masks.json, and the FDA pull that feeds them is reproducible from data/fda_510k.json.

What we deliberately don't do

We don't test masks ourselves, we don't publish a number without a citation, and we don't give medical advice — a 510(k) is a safety-and-equivalence review, not a performance ranking, and nothing here says a mask will or won't do anything for your skin. Where a fact matters — a measured dose, a K-number, a holder, a cleared indication, a recall, an adverse-event report — we quote the source and link the document. If our summary and the source ever disagree, the source wins. We also don't editorialize a mask into or out of the table: the rules are mechanical, and the same rule that benches a $455 unmatched claim would bench anyone else's.

Affiliate links

Some product links are Amazon affiliate links; if you buy through them we may earn a commission at no cost to you. Links appear only on ranked rows — claims matched to a real K-number with a verified price — never on unmatched or Amazon-tier rows, and they never change a verdict, a rank, or a word of a row. The editorial rule is enforced in the site's own code, which is public.

Corrections

If a brand publishes a K-number, a listing, or a price that changes a row — or we got something wrong — the page changes. Every row records the date it was last checked.