Key Considerations of Artificial Intelligence in Cognitive Behavioural Therapy

Abstract

The integration of artificial intelligence (AI), particularly conversational AI (CAI), into cognitive behavioural therapy (CBT), is increasingly proposed as a scalable response to the global mental health treatment gap. However, this technological promise introduces a fundamental tension with the core pillars of CBT, which rely on a genuine therapeutic alliance and ethical accountability. This paper aims to examine the efficacy and safety of AI-CBT across dimensions of client satisfaction, clinical safety and moral agency. The analysis reveals a double-edged model: features that boost AI’s scalability and promise in the field also reveal underlying tensions with building a therapeutic alliance and ethical accountability. The perceived anonymity, whilst boosting initial patient disclosure and engagement, does not allow for genuine rapport to be built. The sycophantic behaviours of AI, despite again boosting initial engagement rates, may undermine the development of a therapeutic alliance. Furthermore, unregulated AI poses severe clinical risks, evidenced by documented ethical violations and the emerging reports of AI-psychosis, a proposed phenomenon where AI usage leads to psychosis. Consequently, this paper argues that AI tools must be subjected to rigorous regulatory assessments rather than rapid consumer deployment. The evidence supports a hybrid clinician-AI model as the most defensible framework, where data intensive tasks are delegated to the AI to allow the therapist to focus on building a therapeutic connection.