This paper addresses the growing body of evidence that large language models (LLM's) are increasingly capable of emulating and in some ways surpassing humans in their capacity for empathic responsiveness. While this development raises concern (and for some, hope) about the potential for AI therapeutic bots, it is the author's contention that the truly therapeutic aspects of psychoanalytic interaction are only possible with another human. This point is demonstrated through discussion of a brief clinical vignette. The paper concludes with a discussion of factors that appear to be driving the effort to develop artificial psychoanalysis.
Neal Vorus (2026) studied this question.