Review of chatbots in urogynecology

Chatbots based on large language models have been rapidly incorporated into many aspects of medicine in a short time despite an incomplete understanding of their capabilities. This review focuses on ways these chatbots have been utilized in urogynecology.

Publications regarding chatbots in urogynecology have centered on patient education, scientific literature review, clinical decision-making, documentation, and research. Several authors have evaluated the ability of chatbots to generate accurate and complete information about prolapse and urinary incontinence. While chatbots can generate accurate information about pelvic floor disorders most of the time, the studies we review indicate that incomplete, misleading, or incorrect information is generated up to 33% of the time. Newer chatbots that are trained for medical applications may help to limit some of these problems. Using chatbots to assist with scientific literature review and research is currently hampered by unpredictable 'hallucinations', where the chatbot may generate information or references that sound plausible but are factually incorrect.

While chatbots are being rapidly integrated into many aspects of medicine, the research evaluating these tools in urogynecology is limited. Publicly available chatbots should only be used for patient education, clinical decision-making, and research with caution.