How we grade a claim

Every entry starts from the strongest available evidence and works down. We prefer systematic reviews and meta-analyses of randomised trials, then individual trials, then guidelines from major bodies, and only then observational work.

The three verdicts

Confidence levels

• High — large, consistent, well-conducted body of evidence; a new study is unlikely to change the verdict.
• Moderate — the direction is reasonably clear, but studies are small, heterogeneous, or limited in scope.
• Low — sparse or weak evidence; treat the verdict as provisional.

Every review is AI, with no human approval

Reviews are not written by an editor, and no human signs them off. They are also not an AI answering from memory. Each claim is run through a live literature search: we query PubMed for recent systematic reviews, meta-analyses, randomised trials and guidelines on that specific intervention and outcome, pull the abstracts, and hand only those retrieved records to the AI reviewer. It is instructed to weigh them by study design and recency, to cite only papers that came back in the search, and to answer “not enough evidence” when the records do not settle the question. Every claim page shows the date of its review and links to every paper used.

What that means for you: an AI can misread a study, miss relevant work, or state a verdict more confidently than the evidence allows. Treat every verdict here as a starting point and read the linked sources yourself. Run your own check any time on the check a claim page.

What this site is not

It is not medical advice, and it cannot account for your history, medication or risk factors. A verdict of “Debunked” means the general claim fails, not that your clinician is wrong about your case. When in doubt, ask them.