Abstract
Artificial intelligence (AI) technologies are advancing rapidly, and their integration to mental health care is being widely explored. Large language model (LLM) chatbots comprise the most recent instantiation of AI and are being developed to provide psychological support, particularly in contexts where traditional (person-to-person) psychotherapy is limited. However, there are many concerns about the quality and the potential risks associated with this emerging technology, including improper and unethical interventions from AI. In this minireview, we examined whether AI-delivered psychotherapy provides good and ethically sound service. This question cannot be addressed without clarifying what constitutes good psychotherapy, and what elements are included in good therapy. Therefore, we first explored these fundamental elements of effective and ethical therapeutic practice, including ethical consideration, practical wisdom, empathy, and theory of mind. We then used this working definition to evaluate the current capabilities of LLM-based systems in mental health context from past research. In conclusion, we suggest that although current LLMs may offer some structured guidance and basic forms of support, they fall short of several essential requirements of competent psychotherapy and therefore cannot reliably replace professional care. To safely integrate LLMs into mental health practice, rigorous governance, ethical oversight, and supervision are crucial.