Background and Aim
Generative AI—and large language models (LLMs) in particular—are opening up new opportunities to extend psychotherapy beyond the consulting room. Their ability to generate context-sensitive, linguistically nuanced responses makes them especially relevant for Mentalization-Based Therapy (MBT), a manualized psychodynamic treatment that strengthens reflective functioning and affect regulation. A central challenge in MBT is sustaining patients' capacity for mentalization outside sessions with the therapist. This work presents Mentora, a prototype tool designed to showcase the feasibility of integrating MBT principles with generative AI to deliver ecological momentary interventions.
Methods and Results
Mentora allows for delivering MBT techniques in a responsive, on-demand fashion through: (1) reflective prompts engineered to promote self-other understanding, (2) feedback on recurrent emotional-linguistic patterns, (3) integration of mood and event journaling into daily routines, and (4) manualized, theory-driven psychoeducational micro-interventions aligned with the MBT theory. The system is configured by therapists, who set treatment objectives and calibrate its functions, while patients regulate scope and intensity. Strong privacy safeguards ensure that patients retain full control over what is shared with therapists. To establish technical adequacy, four LLMs were systematically compared as potential trainers of mentalization. Simulated case scenarios, developed under the supervision of a psychotherapist, illustrate how Mentora can generate reflective material between sessions that later enriches therapeutic work.
Discussion and Conclusions
Mentora demonstrates how generative AI can expand the reach of MBT while preserving the centrality of the therapeutic relationship. By embedding MBT-informed reflection into daily routines, it highlights the potential for scalable, personalised, and ethically grounded augmentation of psychotherapy. Future directions include integrating wearable data (e.g., heart rate variability, sleep quality, stress markers) to enrich ecological momentary interventions and tailor them to patients' mentalizing capacities and well-being.