Editorial methodology
The evidence desk is built to make uncertainty visible and to keep source context attached to every claim.
How AI is used
AI assists with source summarization, claim extraction, structured classification, FAQ drafting and internal consistency checks. AI output is not treated as authority. Source links, evidence labels and risky medical claims require editorial review.
Source selection
We prioritize official regulators, ClinicalTrials.gov, PubMed, peer-reviewed journals, NIH pages and primary company releases. News and social posts can appear as signals, but they are labeled as lower-confidence inputs.
Evidence grading
Evidence labels distinguish human randomized trials, observational human evidence, registered trials, animal or preclinical work, mechanism-only claims, anecdote, regulatory signals, commercial claims and unknown evidence.
Risk labels
Risk labels are editorial warnings about caution level, not personal advice. High-risk, investigational and do-not-self-administer labels trigger human review requirements.
Medical boundaries
We explain source material and questions to ask. We do not tell readers what to take, combine, stop, start, purchase or self-administer.
Correction policy
Corrections are welcomed through the contact page. We prioritize corrections involving regulatory status, source attribution, safety framing and dates.