Large Language Models as Digital Philotherapy: A Structured, Non-Clinical Framework for Low-Intensity Linguistic Support

Global disparities in access to psychological support remain profound, with treatment gaps ranging from 50% to 90% worldwide. Large Language Models (LLMs) are increasingly used for emotional expression and reflective dialogue, yet their role remains conceptually undefined and ethically ambiguous. This paper introduces digital philotherapy, a non-clinical, language-based framework inspired by bibliotherapy, reflective writing, and psychoeducational skills training. We propose a structured, time-limited model—ten sessions with minimal, user-controlled memory—designed to provide continuity of narrative support without constituting psychotherapy. We outline operational boundaries, governance requirements, and regulatory considerations under the EU AI Act, arguing that digital philotherapy can serve as a low-intensity, interim layer of support for populations lacking timely access to mental health services, without entering the domain of medical AI.