Large Language Models as Digital Philotherapy A Low-Intensity, Non-Substitutive Framework for Mental Health Support in Contexts of Limited Access

The principle that there is no health without mental health is widely recognized in global public health, yet access to psychotherapy remains uneven worldwide. Structural, economic, and cultural barriers contribute to substantial treatment gaps, particularly for individuals experiencing mild to moderate psychological distress. At the same time, Large Language Models (LLMs) are increasingly used informally for emotional expression, reflection, and meaning-making. This paper proposes a conceptual framework that positions LLMs not as automated psychotherapy, but as digital philotherapy: a non-clinical, language-based form of low-intensity mental health support inspired by bibliotherapy, reflective writing, and skills-based psychoeducation. Drawing on principles from Dialectical Behavior Therapy (DBT) in a non-therapeutic manner, the framework outlines potential applications, limitations, and ethical safeguards. The paper further discusses regulatory implications under the EU Artificial Intelligence Act, arguing that, when appropriately constrained and transparently governed, such uses fall outside the scope of high-risk medical AI. Digital philotherapy is presented as a complementary and interim support layer for populations with limited or delayed access to professional mental health care, rather than as a substitute for psychotherapy or clinical treatment.