<p>Interest in artificial intelligence (AI) based conversational agents (“AI Chatbots”) for mental health care has grown rapidly in response to recent technological advancements, persistent gaps in access to mental health services, and the fact that people are already using general-purpose AI chatbots for emotional support. Although AI Chatbots may outperform non-AI chatbot digital mental health tools in terms of personalization, accessibility, and perceived empathy, important questions remain, including what constitutes optimal engagement with these tools. We discuss the lessons learned from research on engagement with non-AI digital mental health tools to inform research and implementation of safe, effective and ethical patient-facing AI Chatbots for mental health care. We argue that, similar to evidence-based human-delivered mental health care, AI Chatbots for mental health care should be optimized for engagement with the behavioral targets the interventions were designed to produce rather than engagement with the AI Chatbot itself, how this might contrast with industry incentives for maximizing active and sustained use of these tools, and offer suggestions for clinicians, researchers, and health care administrators to guide the development, evaluation, and implementation of AI Chatbots in mental health care.</p>