Abstract Conversational artificial intelligence (CAI) has gained growing interest in mental health care, including psychotherapy, for its potential to support therapeutic engagement and interventions by enabling context-aware and human-like interactions. One of the key aspects of CAI’s human-like interaction is empathy. However, it is questionable whether (1) the concept of empathy changes when psychotherapy is mediated through CAI, (2) it is meaningful and beneficial to develop CAI simulating empathy, and (3) what ethical challenges arise with CAI’s simulated empathy. To address these questions, the chapter first provides a conceptual, normative, and ethical analysis of various conceptualizations of empathy from the psychological and philosophical literature. Then, it analyses the role of CAI’s simulated aspects of empathy in relation to these conceptualizations, their normative and ethical requirements, as well as the challenges they pose in the context of mental health interventions. Finally, the chapter offers insights and recommendations for designing responsible human–CAI interactions in mental health care.