Corrections policy
A clinical reference is only as good as its error-correction loop. We treat every reported error as a defect in the pipeline, not a one-off typo to patch on the page.
Email corrections@peptidereference.health with:
Reports without a contradicting primary source aren't ignored, but they take longer because the verifier has to find one before the pipeline can act.
Each report is classified into one of four buckets:
Before a correction publishes, the proposed change is independently verified against primary sources. The verifier re-runs PMID, NCT, DOI, and URL resolution; Retraction Watch is consulted for any newly-cited PubMed entry; the affected quality gates are re-evaluated. A correction that fails any gate doesn't ship.
When a substantive change lands on a card, the card's last updated date is bumped and the change is recorded in that card's audit trail. Material corrections (safety-relevant factual errors, regulatory-status changes, tier downgrades) are flagged on the card so a returning reader sees the diff, not just the new version.
We do not silently revise. The correction record exists so that a clinician who acted on a previous version of a card can see what changed and why.
When a single error reveals a pattern (a regex that mis-classified a regulatory status, a synthesis prompt that produced an instruction-leak, a verifier rule that missed a retraction), the fix lives in the pipeline, not just the affected card. The patch is then applied to every card that could have hit the same defect, not only the one that was reported.