What Changed at Vera Health

Vera Health ran 100 questions from the public Real-POCQi dataset of real physician questions through Vera and three frontier models, and reused the study's published answers from OpenEvidence and GPT-5.5. Graded blind by the study's AI judges, Vera was preferred over OpenEvidence on clinical accuracy 80 percent to 7 percent, and Claude Opus 5.5 led on clinical utility and completeness.

Our read

Head to head data on clinical answer tools is scarce, and reusing a published question set makes this one easier to check. All grading was by AI judges, not physicians, and Vera ran the comparison itself.

Vera Health described how it governs answer quality. A jury of AI judges, built on several model families and tuned on clinicians' blinded reviews, grades every change to models, instructions or tools against a fixed set of questions across specialties and countries, and a change ships only if it does at least as well as the current version. The judges also grade a daily sample of live answers, and flagged answers go to a member of the Vera team each morning.

Our read

Vera reports that on 50 held out comparisons its judges matched clinicians 82 percent of the time on accuracy, against 72 percent for a general purpose grader. It is a self reported method, but one buyers can ask competitors to match.