The Gender Gap in AI-Driven Diabetes Care

In this episode, Dr. Gloria Wu — lead author of the 2025 UCSF study "Chatbots and Diabetes: Is There Gender Bias?" — brings her research directly to you as both researcher and host. Published in the Journal of Patient Experience, the study tested four leading AI chatbots — ChatGPT, Gemini, Claude, and DeepSeek — asking each the same question about diabetic vision loss, changing only the patient's gender. The stakes are real: 38.4 million Americans live with diabetes, yet only 19,500 specialists exist to serve them — a gap pushing patients toward AI by default. Dr. Wu walks you through what her team found: all four models communicate at a college reading level, far exceeding the 6th-grade standard recommended for patient education. And each model shows distinct gender-based gaps in the clinical information it provides — with DeepSeek showing the most significant bias against female patients, and the other three models each missing different keywords depending on the patient's gender. Dr. Wu closes with three practical takeaways every patient should know: always cross-check at least two chatbots, ask for plain language, and never treat AI as a substitute for a qualified clinician.