The Audit Method Itself

Nineteen myths have fallen across the previous four files, and every one of them fell the same way. Not by authority, not by reflexive skepticism — by a small set of questions that separate what a claim says from what its evidence shows. This page hands you that method. The sleep industry will keep producing claims faster than any site can audit them; the durable defense is a reader who can run the audit in ninety seconds at the checkout page. Five filters, worked examples, and the reason they work.

Filter One: Who Says It, and Who Pays?

Every claim has a source, and the source's incentives are data. A mattress company's "sleep study" is a marketing asset; a supplement brand's commissioned survey is an advertisement wearing a lab coat; an influencer's "what I take" video has a link in the description. None of that proves falsity — industry can fund good science — but it sets the prior. The professional version of this filter is the conflict-of-interest statement, and its absence from a confident claim is itself information. Ask three questions: who produced this number, what do they sell, and would this claim exist if they sold something else? The site's own answer to this filter is its methodology and editorial standards — published so you can audit the auditor.

Filter Two: Human Evidence, or Something Softer?

Sleep is full of beautiful mechanistic stories — a receptor, a hormone, a pathway — because the biology is genuinely elegant. But the evidence hierarchy is a hierarchy for a reason: what a molecule does in a dish or a mouse sets a plausibility floor, not a ceiling of proof. The audit question is simply "in which rung was this claim tested?" A systematic review of human trials outranks single trials; randomized trials outrank cohorts; cohorts outrank mechanistic inference; and "thousands of satisfied users" is not a rung at all. The glycine story from the site's supplement audit is a clean worked example: real human crossover trials exist, and the honest reading still notes they are few, small, and industry-run — that is filter two applied without cynicism or charity. A speed shortcut for this filter: search the claim plus the word "randomized." If nothing with trial data surfaces in the first page of results, the claim has already told you its rung. And when a study does surface, read the participants line before the conclusions line — a trial in 11 healthy volunteers, 8 rats, or a dish of cells is answering a different question than the one your bedtime is asking.

Filter Three: How Big, Measured How?

"Improves sleep" is not a finding until it says by how much, on what measure, in whom. Minutes of onset? Subjective quality scores? A tracker's estimate? The audit demands numbers with units and a comparison. The classic pattern this filter catches is the tiny effect in a tiny trial reported as a life-changing benefit: "fell asleep 6 minutes faster" becomes "clinically shown to help you sleep" — technically anchored, practically meaningless. The reverse pattern matters too: a genuinely large effect (caffeine cutting an hour of objective sleep) deserves more caution than a small one, not less attention. Run every claim through the units test: percent of what, minutes of which, points on whose scale.

Filter Four: Marketing-Detection

Certain phrases are load-bearing marketing, not information: "restorative blend," "clinically proven" with no trial named, "doctors recommend" with no organization, "the hidden answer," anything that frames sleep as something to optimize with a purchase. The vocabulary tracks whatever sells — gadgets one decade, supplements the next — but the grammar is stable: urgency, exclusivity, and a scientific veneer thin enough to see through on the second read. A useful habit: highlight every adjective in a claim and ask what evidence each one carries. "Advanced," "smart," "natural," and "drug-free" each carry none. This filter alone dispatches most sleep-industry content before the science even loads.

Filter Five: Does It Fit the Machine?

Sleep runs on two processes — homeostatic pressure that builds with hours awake, and a circadian clock that times the cycle — and claims that violate the machine announce themselves. "Reset your sleep in one night" ignores that pressure and clock are physiological systems, not settings. "Train yourself to need four hours" contradicts the biology the adaptation-illusion research measured. This filter requires the most background knowledge, but a serviceable version fits in one sentence: when a claim promises sleep improvements without touching sleep pressure, timing, or a medically plausible pathway, it is selling something the machine has no input for. The two-process model itself is covered honestly in Circadian Rhythm 101. A bonus of this filter: it also catches under-claims. When a genuine lever — morning light, a consistent wake time, caffeine cutoff — sounds boring next to a gadget, the plausibility check reminds you which one actually operates the machine. The most evidence-backed sleep interventions are aggressively unglamorous, and that itself is a filter result: the machine responds to inputs it recognizes, and it does not recognize marketing.

The Mouth-Tape Worked Example

Run the trend end-to-end. Filter one: the loudest mouth-tape voices sell tape or content. Filter two: human evidence is nearly absent — no controlled trials showing sleep improvement; a small snoring-and-mouth-taping study gets cited far beyond its scope. Filter three: benefits are never quantified; "better sleep" carries no number. Filter four: "the ancient insider tip," "what elite performers do" — the grammar is pure. Filter five: nasal breathing is real physiology, but the tape does nothing about why the mouth opens — and in someone with undiagnosed apnea or nasal obstruction, forcing the route shut is a genuine hazard, not a neutral experiment. Verdict: fails four filters, and the fifth raises a safety flag demanding a clinician's input — which is exactly why ⚠️ cautions belong to the method, not just the verdicts. Suspicion of claims must never curdle into suspicion of symptoms.

Why the Method Beats Memorized Verdicts

This topic could have been twenty verdicts to memorize; it ends instead with the tribunal that produced them, because the verdicts expire and the method doesn't. Next year's sleep trend — a wearable, a drink, a protocol with a name and an app — will not be on any list. It will have a source with something to sell, a rung (or absence) on the evidence hierarchy, an effect size hiding in a percentage, marketing grammar, and a fit or misfit with two-process physiology. Those five attributes are as detectable in a product that launches tomorrow as in the ones this pillar audited yesterday. That is the quiet promise of the whole file series: not a list of answers, a working habit of questions.

Questions, Answered Briefly

• ⏱️ "Ninety seconds per claim — really?" The first two filters are near-instant (source and study-rung take one scroll); the full run needs the claim to actually name its evidence, and most never do.
• 🤝 "What if a funded study is still good?" Then it survives the filter — funding sets the prior, not the verdict. Read the methods; discount the framing.
• 🔗 "Where do I check a study's funding?" The conflict-of-interest or funding statement near the end of any legitimate paper — one minute of scrolling; its absence from a summary you're reading is the finding.
• 🧠 "Do I need the two-process model for filter five?" The one-sentence version suffices: real levers touch pressure (time awake, sleep hours) or timing (light, schedule) — or a clinician-plausible pathway. Everything else is decoration.

Related Topics