Hallucinated Citations and Clinical Risks
AI Chatbots Fail 50% of Medical Queries in New Study
A BMJ Open audit reveals that nearly half of AI chatbot responses to medical queries are inaccurate or incomplete.
A close-up of a person's hand holding a smartphone displaying a chat interface with a red warning icon, set against a blurred medical office background.
Photo: Avantgarde News
A study in BMJ Open found five AI chatbots provide poor medical advice half the time [1][2]. Researchers evaluated responses to health queries and found 50% were inaccurate or incomplete [1]. These findings raise concerns about the reliability of artificial intelligence for health information [3]. The audit also highlighted frequent hallucinations where the chatbots invented scientific citations [1]. These tools often provided links or references to studies that do not exist [2]. Experts warn that these systems are not suitable for clinical decision-making or self-diagnosis [3].
Editorial notes
Transparency note
Drafted with LLM; human-edited
- AI assisted
- Yes
- Human review
- Yes
- Last updated
Risk assessment
The topic involves medical misinformation risks and public health safety.
Sources
Related stories
View all
Topics
About the author
Avantgarde News Desk covers hallucinated citations and clinical risks and editorial analysis for Avantgarde News.