Large Language Models at Population Scale: A Survey and Taxonomy of Public Health Applications

This paper presents the first comprehensive survey of LLMs in public health through a two-dimensional taxonomy, which maps six core public health tasks against five functional LLM roles. Our analysis reveals a critical gap: while current LLMs excel at extracting health signals from text, they struggle with the complex reasoning required for population-level challenges. We argue that LLMs must evolve from individual-level clinical assistants into societal-scale reasoning engines capable of modeling disease dynamics, simulating policy interventions, and ensuring health equity on a global scale.