Effect of ChatGPT Feedback on the Outcomes of a Web‐Based Self‐Care Support Tool Based on Solution‐Focused Brief Therapy

Large language models can provide personalized feedback within web-based self-care support tools. This study investigated the impact of ChatGPT feedback, which provided advice aimed at improving the concreteness and realism of the user's goals, in a self-care support tool based on solution-focused brief therapy. Participants were randomly assigned to the ChatGPT group (n = 92) or the control group (n = 82). The tool comprised eight components: problem definition, miracle question, goal setting, machine learning model-based feedback on goals, ChatGPT feedback (ChatGPT group), goal re-setting, exception question, and ideas for problem solving. Our findings indicated that the self-care support tool was effective, improving solution building, positive affect, the level of realization of the ideal life, and problem severity in both groups. The additional effect of incorporating ChatGPT feedback was demonstrated only for problem severity. These results suggest that self-care support tools are effective, with ChatGPT feedback providing an added benefit for problem severity.