Therapeutic Alliance With Artificial Intelligence: A Critical Analysis of Digital Relationship Formation

Abstract

Drawing on therapeutic alliance research, phenomenological philosophy, and empirical evidence from artificial intelligence (AI) mental health applications, this theoretical paper proceeds in three parts to establish the argument that a genuine therapeutic alliance cannot be formed with current AI chatbots. First, it reviews evidence on AI chatbot use in mental health, including both promises and adverse effects. Second, it analyzes therapeutic alliance measurement tools, including the Working Alliance Inventory (WAI) and its digital adaptations. Third, it examines mechanisms such as anthropomorphic projection and sycophancy through which AI may trigger alliance indicators without constituting a genuine relationship. Throughout, the paper applies a phenomenological framework positing that authentic therapeutic relationships require consciousness, embodiment, shared intentionality, and existential commonality. The paper concludes that current AI chatbots cannot fulfill the ontological requirements for an authentic therapeutic alliance. Implications are discussed for future research and guidance on how to incorporate the findings into clinical work.