OpenEvidence tops rheumatology AI benchmark with 91.2% guideline agreement across 60 scenarios
benedictneo · x · 2026-11-09
A study presented at the ACR plenary tested five AI platforms on 60 scenarios built from ACR and EULAR guidelines; OpenEvidence ranked first with 91.2% agreement. Rheumatology was chosen deliberately as a hard test—decisions hinge on patient-level details and rapidly changing guidelines. The study notes that even convincing answers need checking, which is why OpenEvidence shows sources with every answer. The company calls for more independent evaluations like this.