Auditing Biomedical Literature with Blinded AI Peer Review: Where Peer Review Works, Where It Fails, and How to Strengthen It

BARS (Blinded AI Review System) produced 1,874 blind reviews and achieved 100% pass rates on verified sound controls while detecting 77%–91% of internally flawed papers.

• Why it matters: Biomedical publishing has expanded far faster than human peer review, and unblinded language models can’t audit literature objectively due to memorizing famous authors and journals.
• What they did: BARS strips titles, author identities, institutions, journal headers, and dates, then grades full texts on a 10-point rubric; it also audited 320 open-access articles (2020–2025) and 915 date-blinded papers (1991–2025).
• The result: Core rigor failures were common (54% scored <5.0/10; 41% excluding megajournals), scores declined from 7.6/10 to 5.8/10 over time, and BARS recalled 98% of unresolved human referee critiques.

AI summary of the abstract · may contain errors · not medical advice · read the paper.