Metaphorical Comics at the Interface of AI and Mental Health: Toward Patient-Centered Visual Expression

This paper explores the potential of large language models (LLMs) and generative AI to support doctor–patient communication by helping individuals with mental health challenges express their illness experiences through metaphorical comics. Comics and visual metaphors offer an accessible way for individuals who struggle with verbal or conventional visual expression to communicate emotions, inner states, and lived experiences. Previous research has largely focused on non-AI-based expressive approaches. Yet these methods remain limited by patients’ drawing abilities and the need for continuous support from trained facilitators. Based on a small study with ten participants, we found that although generative AI offers new possibilities—enabling patients who cannot draw to externalize internal states —current systems still struggle to capture metaphor-based emotional expression and maintain coherent narratives across panels. Building on these observations, this paper outlines technological pathways that may enhance the feasibility of AI-supported visual metaphor generation in graphic medicine. These include constructing a domain-specific repository of metaphorical elements commonly found in mental health narratives and using retrieval-augmented generation (RAG) to connect patients’ lived accounts with these elements, enabling LLMs to produce context-sensitive prompts informed by art-therapy practice. Finally, the approach can be further strengthened by integrating such metaphor-informed text workflows with sketch-conditioned text-to-image models to improve narrative coherence and emotional resonance. Taken together, these pathways offer a conceptual framework for how generative AI might be developed to support patient-centered expressive practices and to expand methodological approaches within comics-as-research, while also highlighting the technical and interpretive challenges that remain.