The tracked audit surface
Every review of a website is a sample: somebody opens some pages and reports what they happened to find. Sampling has no end state, because there is always another page nobody opened. This page is the other approach. It lists the classes of claim this site tracks and gives each one of three states, recomputed on every deploy rather than asserted.
One limit stated plainly, because it is the difference between a useful figure and an overstated one. The classes below are a registry maintained by hand. This page does not discover every possible species of claim on the site and prove each has a gate; it reports the state of the classes that are being tracked. So the last number does not mean nothing on this site is unchecked. It means every class currently in the registry has an audit path.
18 tracked classes are verified by a gate on every build, 8 carry declared and ratcheted debt, and 0 have no audit path. Adding a class to the registry is how this page gets more honest, not less, so the work is to keep finding classes rather than to keep the last number at zero.
A gate checks every unit of these classes and a failure stops the deploy.
Every letter declares an endpoint The outcome a trial measures to decide whether the treatment worked, such as heart attacks, blood pressure or a symptom score. type and none exceeds its ceiling. Checked by: G29, G32.
Every one carries an adversarial judgment; A, B, C and D all enforced. Checked by: G49.
Route 1 No human randomized trial has been run, so the record rests on animal or mechanistic evidence only. One of the two routes to the lowest letter on EBL's scale. shows an absence check; Route 2 Human randomized trials have been run and did not show benefit on the stated outcome. One of the two routes to the lowest letter on EBL's scale. names the null or harm trial. Checked by: G70.
Every authored citation and per-use split reaches the page it grades. Checked by: G72.
Re-checked against PubMed every build for retraction and expressions of concern (counted by refcount.py, the owner of this figure). Checked by: G5, G14, G25, G37, G48.
Disclaimer, landmarks, accessible names, meta description, share card, valid inline script. Checked by: G15, G19, G22, G42, G64, G75.
Vocabulary constrained, no row claims more than its tier allows, a row that disowns its own claim says what it is instead, and no row without a human trial asserts a benefit in its repletion box. Checked by: G23, G35, G111, G126.
57 of 57 medication records publish a page. The exception is deliberate: muscle_relaxants is a string-only record, its page is undeclared so the build prunes it, and _redirects sends the URL to the drug index. A count of 57 against 57 is the intended state, not a missing page. Checked by: G18, G88.
Every one states whether it rests on a study or on pharmacology. Checked by: G40, G47.
Each carries an evidence category and a source; no page prints a target the evidence lacks. Checked by: G39, G71.
Scanned in the published tree after pruning, so a corrected claim cannot return. Checked by: G69.
Each row's own text states the drug has not been shown to deplete the nutrient, so there is no citation to give. Allowed only on mechanistic or traditional rows, never critical. Checked by: G35.
Every one of the 136 rows states where its figure came from, or states that no dose is given and why. 72 did not on the morning of September 2026. The counter stays in the debt ratchet at zero, so the next dose written without a source fails the build rather than joining a backlog. Checked by: G73, G126.
Every page offering a way to buy carries the commercial disclosure. The two facts were both true and nothing tied them together until G76. Checked by: G76.
Every page offers the same destinations, or is declared by name with a reason. gate_nav checks a nav exists; G77 checks they agree. Checked by: G77.
Every /go/ link resolves to a rule in _redirects, and no specific rule sits below a wildcard that already swallows it. Checked by: G78.
Every `avoid` claim declares whether it rests on a study, a mechanism, absent data, or nothing prominent, and the basis renders beside the claim. Checked by: G79.
Every letter a hub publishes is one its record holds, and every record it links to exists. The earlier entry here judged hubs by an entity-page template and was measuring the wrong thing; five stale letters were the real defect and are fixed. Checked by: G80.
The work is not done. The number is known, dated, published, and ratcheted so it can fall but cannot rise.
Ratcheted: may fall, may not rise. Checked by: G73.
Ratcheted. Each falls when a dated PubMed search is recorded on it. Checked by: G73, G184.
Ratcheted. Stating the basis is the floor, not the standard. Checked by: G73.
Ratcheted. Published on the Evidence Integrity page. Checked by: G73.
Ratcheted. Their evidence renders inside other pages. Checked by: G73.
Coverage ratchet: may rise, may not fall. Checked by: G63, G73.
Coverage ratchet: may rise, may not fall. Checked by: G63, G73.
Ratcheted. Assigning a basis to these is a judgment about safety content and is deliberately left to a person rather than guessed. Checked by: G73, G79.
No gate reads these. They are listed because an unmeasured surface is where every defect on this site has been found, and naming a class is what turns an endless stream of corrections into a finite list.
Educational use only · Not medical advice · EBL grades published human evidence and does not recommend treatment. · Privacy · Disclosure · Terms · Accessibility · Clinical trials
Evidence Based Longevity, Inc. is owner-operated and edited by Arnold Sandlow, D.C., a California-licensed chiropractor. Evidence Based Longevity publishes in English only · Machine translations are not produced or reviewed by EBL for accuracy
© Evidence Based Longevity, Inc. All rights reserved