1.2 Problem statement and relevance
Empirical research into psychotherapy shows that psychotherapy is not primarily effective because of any specific method or technique; rather, healing occurs through the quality of the therapeutic relationship within the interpersonal interaction. The integration of artificial intelligence (AI) into psychotherapy therefore raises the pressing question of whether this therapeutic relationship must be human, or whether AI might in future not only support therapists but permanently replace them.
Research to date has focused primarily on functional aspects and symptom reduction: recent studies show that chatbots (such as Woebot) are indeed successful in providing support for mental health issues and are sometimes even perceived as more empathetic than human professionals in user reviews.
Given current problems in healthcare policy and a massive shortage of qualified staff within the psychosocial care system, it seems tempting to outsource valuable therapeutic relationship-building work to such systems. One might succumb to the belief that empathy can be technically simulated in order to counteract the shortage of practitioners. However, this ‘delusion of feasibility’ (in the sense described by Fuchs and Rosa) threatens to lead to a creeping alienation and de-embodiment of psychotherapy. This raises the key question of the extent to which these systems are structurally capable of facilitating a genuine therapeutic encounter in the sense of a bodily ‘I-You’ relationship.
This question is already being debated within the Gestalt therapy community. At the 2025 DVG conference in Freiburg, a live session between a model client and ChatGPT was conducted in front of around seventy participants and subsequently discussed in plenary. Bauer (2026) documents the experiment and discussion in the journal *Gestalttherapie*. His conclusion identifies what the machine lacks: ‘no body, no biography of its own, no responsibility, no genuine reciprocity’ (Bauer, 2026, p. 89). This list is accurate. However, it merely states that something is missing, not why it must be missing. This is precisely where this paper comes in.