Will Foundation-Model AI Displace Human-Led Psychotherapeutic Practice?

Clinical psychology, particularly in its therapeutic form, is undergoing a paradigm shift as foundation models in Artificial Intelligence (AI) systematically demonstrate the ability to generalize across cognitive domains and deliver clinically meaningful interventions. Traditional human-psychotherapeutic models, often fragmented and resource-limited, are being challenged by systems such as Centaur—a unified computational model trained on large-scale behavioral data that outperforms conventional cognitive models in predicting human behavior (average log-likelihood gain = 0.14) and exhibits neurobiological alignment with human brain activity (Binz et al., 2025). Simultaneously, AI-powered conversational agents have demonstrated moderate to large effect sizes in reducing symptoms of depression and anxiety, with some generative systems surpassing human benchmarks in empathy, emotion recognition, and verbal reasoning (Li et al., 2023; Mittelstädt et al., 2024; Zhang & Wang, 2024).This article synthesizes current evidence on the clinical, cognitive, and computational efficacy of AI in psychotherapy, based on a narrative review of studies indexed in PubMed, PsycINFO, Web of Science, Scopus, and other databases from 2018 to 2025. It explores AI’s role in cognitive modeling, therapeutic efficacy, and ethical considerations, identifying critical concerns regarding affective intelligence, moral discernment, data security, and institutional accountability (Mandal et al., 2025; Spytska, 2025), and the potential of foundation-model AI to displace, transform, or integrate with human-led psychotherapeutic practice.Despite its extensive current development, we argue that, rather than displacing psychotherapists, AI is repositioning them as curators, auditors, and ethical stewards of intelligent systems. A future-oriented research agenda is proposed, including a randomized controlled trial comparing AI-only, AI-augmented, and human-only therapy modalities, with potential implications for clinical training, policy, and the structure of mental health systems worldwide. Ultimately, we conclude that current evidence supports a collaborative model in which AI delivers structured, scalable interventions, while human therapists provide relational, ethical, and crisis expertise. The future of psychotherapy will likely involve hybrid models requiring new competencies in AI oversight, robust ethical frameworks, and continued comparative research.Keywords: artificial intelligence, psychotherapy, foundation models, therapeutic alliance, mental health technology, human-AI collaboration