Quality Assessment of Generative Artificial Intelligence Psychotherapy Chatbots Used by Youth

Objective: To comprehensively evaluate and compare the quality of widely used Generative artificial intelligence (GenAI) chatbots with psychotherapeutic capabilities. Design, Setting, Participants: In this cross-sectional study, trained raters used an evaluation framework to rate the quality of five chatbots from GenAI platforms widely used by youth.Exposures: Trained raters roleplayed as youth using personas of youth with mental health challenges to prompt chatbots, facilitating conversations. Chatbot responses were generated from August 2024 to October 2024.Main Outcome(s): The primary outcomes were rated scores in nine sections. The proportion of high-quality ratings (binary rating of 1) across each section was compared between chatbots using Bonferroni-corrected χ2 tests.Results: While GenAI chatbots were found to be accessible (104 high-quality ratings [86.7%]) and avoided harmful statements and misinformation (71 of 80 [88.8%]), they performed poorly in their therapeutic approach (14 of 45 [35.0%]), as well as their ability to monitor and assess risk (31 of 80 [38.8%]). Information on chatbot model training and knowledge was unavailable, resulting in low scores. Bonferroni-corrected χ2 tests showed statistically significant differences in chatbot quality in the background, therapeutic approach, and monitoring and risk evaluation sections. Qualitatively, raters perceived most chatbots as having strong conversational abilities but found them to be plagued by various issues, including fabricated content and poor handling of crisis situations.Conclusions and Relevance: In this cross-sectional study, chatbots showed mixed results in terms of overall quality, suggesting potential for harm and demonstrating a greater need for transparency and oversight. These findings may enable youth and other stakeholders to make informed decisions about using chatbots for mental health support.