Abstract
Global mental health systems face a persistent gap between the prevalence of psychological distress and the availability of clinical resources. Artificial intelligence (AI), particularly conversational agents and wellness applications, may help address this gap, yet the field lacks a unified framework explaining how these tools engage users and support sustained psychological well-being. Drawing on psychotherapy integration theory, dynamic systems theory, value-sensitive design, and self-determination theory, this review introduces the AI-DEA (Artificial Intelligence, Design, Engagement, and Adaptation) framework. The model comprises five interacting nodes (chatbot design, AI algorithmic model, engagement, adaptive use, and psychological outcomes), linked through nine pathways. We integrate recent evidence from randomised controlled trials, clinical alignment studies, and automated safety benchmarks to situate the framework within current findings on chatbot effectiveness, crisis-risk calibration, and feedback dynamics. AI-DEA also identifies key moderators, including AI attitudes and individual differences, and incorporates ethical considerations related to safety, regulation, and developmental safeguards for adolescent users. The framework offers a theoretically grounded, empirically informed roadmap for designing, evaluating, and responsibly deploying next-generation AI wellness platforms.