Signal Whose doctor does the AI recommend? An algorithm audit of reputation and demographic signals in large language model-assisted physician choice

Summary

A prespecified randomized algorithm audit examines how large language model assistants recommend physicians to patients. The authors note that patients increasingly ask such assistants which doctor to see, turning these systems into quiet 'AI infomediaries' that decide at scale which physicians get exposure. The study tested seven models, six open-weight and one from OpenAI's GPT family, to causally establish whether factors such as reputation and demographic signals actually move recommendations. It positions itself as an algorithm audit of a consumer-facing AI application with direct implications for healthcare access. The researchers suggest this kind of check could serve as a template for auditing fairness in similar systems going forward.