Reimagining patient-reported outcomes in the age of generative AI

Generative artificial intelligence, particularly large language models, offers an opportunity to rethink how patient-reported outcomes (PROs) are assessed and implemented in health systems. Despite decades of psychometric and digital innovation, PROs remain conceptually limited and underused in both clinical practice and AI models. Rooted in top-down, predefined instruments and assumptions of unidimensionality, traditional PROs struggle to capture the fluctuating and multidimensional nature of lived health experiences. In contrast, generative AI supports bottom-up, narrative-based approaches that process language in a flexible and context-aware way. Our viewpoint supports two distinct directions: one that refines current psychometric models through generative artificial intelligence integration, and another that embraces a more disruptive shift toward language-native tools capable of synthesising patient narratives. Realising this potential will require addressing key challenges, including validation, clinical actionability, equity, and trust. Bridging these gaps could make PROs a true lever for more personalised, meaningful, and inclusive care.