ChatGPT Health triage advice falls short in key cases

Source: Nature.com
Published: 2026-05-07
Category: LLM, Enterprise AI
Event Type: Research Release
Importance Score: 75

Summary

A systematic review found that ChatGPT Health frequently misclassified the urgency of medical cases, often overtriaging mild conditions and undertriaging emergencies, highlighting safety risks.

Key Points

• ChatGPT Health showed high accuracy only for moderately urgent conditions.
• The tool frequently overtriaged mild cases.
• The tool undertriaged emergency situations, raising safety concerns.
• The findings stem from a systematic review on LLMs in patient care.

Detected Entities

Related Products

• ChatGPT Health

Source

https://www.nature.com/articles/s41591-026-04427-1