Corrections
A site with no corrections page is not a site with no errors. This is the log of substantive corrections to published health claims — what was wrong, and what replaced it.
#How to report one
Use the feedback link on any page, or write to [email protected]. The most useful report names the page, quotes the sentence, and says what the source actually shows. You do not need to be polite about it and you do not need a credential.
Claims are corrected at their source, so a fix propagates to every page that repeats it — usually within a day.
#Corrections made
This log begins on 27 July 2026, when the site was first audited end to end. It covers corrections to published claims; routine editing and code changes are not listed.
#A trial figure attributed to the wrong study
The clenbuterol page repeated a widely quoted "+39% fat oxidation" and attached it to the randomised controlled trial in healthy humans. The controlled trial reports no such thing — it measured a fat-mass change of 0.00 kg. The 39% comes from a separate open-label, uncontrolled study of six men, measured 140 minutes after a single dose. The number is now shown with its real source and its real design, next to the controlled result it was being used to dress up.
#Animal data was earning human evidence ratings
The star rating was taking the highest evidence grade on a compound regardless of species, so compounds with strong rodent data and thin human data could display a high human-evidence rating. Rapamycin was the clearest case at five stars; it now shows three, which is what its human evidence supports. Animal-only compounds are capped at two stars and labelled, and the build refuses to publish if that is ever violated again.
#Prescription medicines described as available over the counter
The site was reading a regulatory approval badge as if it were a supply classification. Seven prescription-only medicines consequently displayed as available from ordinary retailers, while a common vitamin displayed as prescription-only. Approval and availability are now separate fields and are stated separately.
#A protocol prescribing the movement it told you to avoid
The rotator-cuff protocol listed an overhead press, on a page whose own advice was to avoid overhead loading. The exercise tags behind it had been assigned by muscle group rather than by function. Tags are now derived from the movement itself, and a protocol that would prescribe a movement it contraindicates fails the build.
#A dose calculator returning milligrams where grams were correct
One calculator produced a result a thousand times too small. Every calculator now carries a machine-readable maximum and is checked at build time.
#Anxiety listing no first-line treatment
The anxiety protocol discussed supplements without naming cognitive behavioural therapy or SSRIs — the treatments with the strongest evidence for the condition. Omitting them made the supplement evidence look stronger by comparison than it is. Both are now named.
#A bone-health claim stated backwards
The menopause protocol inverted the direction of an effect on bone density. Corrected.
#A triage questionnaire with no cardiac red flag
The neck-and-shoulder questionnaire screened for nerve and structural warning signs but not for the cardiac presentation its own escalation text opened with. It now screens for it, and withholds the protocol in favour of emergency advice when it fires.
#A trial cited to the wrong journal
A phenibut trial was credited to Drug Testing and Analysis; it was published in Pharmacopsychiatry. The page body had the right journal in the text while the citation had the wrong one — the page was contradicting itself.
#Microdosing conclusions drawn from full-dose trials
The psilocybin entry presented five trials together under a microdosing heading. Four used full 10–25 mg doses with psychological support; the one genuine microdosing trial — self-blinded, placebo-controlled, 191 completers — found the microdose and placebo groups improved equally. Nothing on the page distinguished them, so four strong full-dose results read as support for microdosing. The entry now states the distinction before the results.
#Compounds ranked against the wrong goals
72 of 387 goal assignments were wrong, including a wakefulness-promoting drug ranked first under "Sleep Better". The taxonomy was rebuilt and is now checked at build time.
#Reviewer credentials that were not real
This is the worst one, and it is published here because a corrections log that leaves out its worst entry is not a corrections log.
The site displayed community protocols attributed to a "verified physiotherapist", alongside other named contributor accounts. None of those people existed. The accounts and their content were seeded demonstration data, and the verification badge was not backed by any check — there was no verification process behind it at all. Everything in that cluster has been deleted.
The wording around it went too. The site said verified clinicians stewarded protocols in their field, that a verified dietitian checked corrections to food data, that a verified physiotherapist could attach anatomy models, and that a verified expert would build a requested protocol. None of that was happening. There are no verified professionals here, no programme that verifies anyone, and no clinician has reviewed this content — which is what the line at the foot of every protocol page has always said, and is now the only thing the site says on the subject.
#Citations audited end to end
Every citation in the corpus was checked against its record. Across two passes, 244 citations were bound to verified sources, 58 claims were corrected to match what the source actually reports, and 8 were rejected outright and removed. Where no source could be found for a claim, the claim was removed rather than left standing.