Effectiveness of Topic-Based Chatbots on Mental Health Self-Care and Mental Well-Being: Randomized Controlled Trial (Preprint)

BACKGROUND The global surge in mental health challenges has placed unprecedented strain on health care systems, highlighting the need for scalable interventions to promote mental health self-care. Chatbots have emerged as promising tools by providing accessible, evidence-based support. While chatbots have shown promise in delivering mental health interventions, most studies have only focused on clinical populations and symptom reduction, leaving a critical gap in understanding their preventive potential for self-care and mental health literacy in the general population. OBJECTIVE This study evaluated the effectiveness of a rule-based, topic-specific chatbot intervention in improving self-care efficacy, mental health literacy, self-care intention, self-care behaviors, and mental well-being immediately after 10 days and 1 month of its use. METHODS A 2-arm, assessor-blinded randomized controlled trial was conducted. A total of 285 participants were randomly assigned to the chatbot intervention group (n=140) and a waitlist control group (n=145). The chatbot intervention consisted of 10 topic-specific sessions targeting stress management, emotion regulation, and value clarification, delivered over 10 days with a 7-day free-access period. Primary outcomes included self-care self-efficacy, behavioral intentions, self-care behaviors, and mental health literacy. Secondary outcomes included depressive symptoms, anxiety symptoms, and mental well-being. Assessments were self-administered on the web at baseline, 10 days after the intervention, and at a 1-month follow-up. All outcomes were analyzed using linear mixed models with an intention-to-treat approach, and effect sizes were calculated using Cohen d. RESULTS Participants in the chatbot group demonstrated significantly greater improvements in behavioral intentions (F2,379.74=15.02; P<.001) and mental health literacy (F2,423.57=4.27; P=.02) compared to the control group. The chatbots were also able to bring significant improvement in self-care behaviors (Cohen d=0.36, 95% CI 0.08-0.30; P<.001), mindfulness (Cohen d=0.37, 95% CI 0.14-0.38; P<.001), depressive symptoms (Cohen d=–0.26, 95% CI –1.77 to –0.26; P=.004), overall well-being (Cohen d=0.22, 95% CI 0.02-0.42; P=.02), and positive emotions (Cohen d=0.28, 95% CI 0.08-0.54; P=.004) after 10 days. However, these improvements did not differ significantly at 1 month when compared to the waitlist control group. Adherence was higher among participants who received push notifications (t138=–4.91; P<.001). CONCLUSIONS This study highlights the potential of rule-based chatbots in promoting mental health literacy and fostering short-term self-care intentions. However, the lack of sustained effects points to the necessary improvements required in chatbot design, including greater personalization and interactive features to enhance self-efficacy and long-term mental health outcomes. Future research should explore hybrid approaches that combine rule-based and generative artificial intelligence systems to optimize intervention effectiveness. CLINICALTRIAL ClinicalTrials.gov NCT05694507; https://clinicaltrials.gov/ct2/show/NCT05694507