Boston Children's Hospital deployed an AI system it calls a 'co-pilot geneticist' that analyzes genetic data, patient symptom descriptions, and global medical literature to surface candidate diagnoses for children with rare, previously undiagnosed diseases. As of mid-2026, the system has contributed to more than 40 confirmed diagnoses that specialists had not been able to make through conventional analysis.

Researchers at the hospital's Manton Center for Orphan Disease Research used OpenAI's o3 Deep Research reasoning model to analyze de-identified clinical and genomic information from 376 previously analyzed cases that remained unsolved. The model received clinicians' notes, patient symptom descriptions, and filtered gene candidate lists as input, and produced evidence-linked candidate explanations for specialist review. Human researchers reviewed all outputs; a finding counted as a confirmed diagnosis only after expert review, additional testing, laboratory confirmation, and return of results by a clinical team. The study was published in NEJM AI. As a result, 18 new diagnoses were confirmed in an initial published study, with the hospital reporting more than 40 total diagnoses to date across its broader AI-assisted program.