Psychological Impact of AI-Mediated Therapy on Treatment Outcomes: A Mixed-Methods Evaluation

The rapid expansion of artificial intelligence (AI) in mental health care has introduced new forms of therapeutic support, yet its psychological impact on treatment processes and clinical outcomes remains insufficiently understood. This study investigates the effectiveness of AI-mediated therapy and examines the psychological mechanisms through which AI influences symptom change. Using a mixed-methods design, participants were assigned to AI-mediated therapy, human-delivered therapy, or a blended model integrating AI support alongside clinician sessions. Quantitative outcomes were assessed using validated clinical measures, including the PHQ-9, GAD-7, Working Alliance Inventory, and engagement metrics such as session adherence and module completion. Qualitative interviews with a purposive subsample explored perceptions of empathy, trust, usability, and therapeutic alliance within AI-based interactions.

Preliminary findings indicate that AI-mediated therapy produces clinically meaningful reductions in anxiety and depressive symptoms, with outcomes comparable to human-only therapy. Mediation analyses suggest that therapeutic alliance and engagement partially explain this improvement, although alliance scores were consistently lower in the AI-only condition. Moderator analyses further reveal that age, baseline severity, and technology familiarity influence both engagement and outcome trajectories. Qualitative themes highlight the importance of transparency, conversational naturalness, and perceived emotional responsiveness for patient acceptance.

These findings demonstrate that AI-mediated therapy can be an effective component of mental health care, particularly when integrated with human support, while emphasising the need for psychologically informed AI design.