AI Driven Mental Health Chatbots: Perceived Empathy, User Satisfaction and Treatment Outcomes

In this study, a quantitative 2x2 factorial experimental design is used to explore how AI transparency, theory of change (ToC), therapy style of advice, AI acceptance rate and type of mental health issue influence user perceptions of AI-driven mental health chatbots. As artificial intelligence (AI) continues to evolve, its potential role in online mental health therapy is gaining increasing interest. In this study, a quantitative 2x2 factorial experimental design is used to explore how AI transparency, theory of change (ToC), therapy style of advice, AI acceptance rate and type of mental health issue influence user perceptions of AI-driven mental health chatbots. Using a mixed-methods approach that combines quantitative analysis with sentiment and emotional text mining, the research examines how these variables shape user experiences in terms of perceived empathy, satisfaction and treatment outcomes. The findings reveal that participants who are aware they are interacting with AI tend to report more positive experiences, particularly when an emotional ToC is employed. Furthermore, emotional advice styles elicit deeper emotional engagement, while rational advice is associated with more positive sentiment. Additionally, the emotional tone and conversational dynamics vary by discussion topic, with depression-related conversations showing greater emotional intensity. These insights underline the importance of aligning chatbot communication styles with individual user expectations and emotional needs, offering implications for the design of more personalised mental health technologies.