PATHFinder AI creates tailored prenatal care plans with GPT-5.2

Researchers from the University of Michigan and other institutions have developed PATHFinder Agent (Planner for Appropriate Tailored Healthcare), an end-to-end conversational AI system designed to deliver personalized prenatal care. The system aligns with the American College of Obstetricians and Gynecologists (ACOG) new PATH guidelines, which advocate for tailored prenatal care plans. PATHFinder operates through a four-stage workflow: patient intake, dynamic interaction, plan synthesis, and clinician oversight. It uses large language models (LLMs) to gather patient health and social context via structured dialogue, then curates individualized care plans and surfaces community resources from Michigan 211.

In evaluations using expert-curated rubrics across five clinical dimensions, GPT-5.2 achieved the highest average score at 77.6%, though the authors identified key gaps in antenatal testing recommendations. The system, accepted as a demo at ACM Interactive Health 2026, represents a leap toward AI-assisted, personalized prenatal care at scale. Future work includes human participant studies and randomized controlled trials to validate clinical effectiveness. This approach promises to reduce physician burden and improve outcomes by ensuring each patient receives a plan tailored to their unique needs and local resources.