Patients often arrive at short doctor visits unsure what to expect or ask. General-purpose AI chatbots, specifically large language models (LLMs), are widely used for health prep, but their impact was untested.
We built a structured LLM visit preparation agent that guided patients through context gathering, expectation setting, question building, and summary. 802 U.S. adults in the Google Health app were randomly assigned this agent or a basic control chatbot.
The structured agent significantly outperformed the control. Blind clinician reviews rated it more comprehensive and aligned to patients' health context. Patient surveys showed higher preparedness, confidence, and satisfaction. This shows LLMs can safely prepare patients when following a structured workflow.