AI matches clinicians on diabetic foot-care guidance

AI-generated guidance on diabetic foot-ulcer self-care matched clinicians for accuracy, understandability and actionability in a comparison covering 30 common patient questions.

Foot ulcers can become serious complications of diabetes, and clear advice on day-to-day care can help people reduce risk. Clinical teams, however, need education materials that are both safe and easy to use.

Researchers at Shanghai’s Second Affiliated Hospital of Naval Medical University compared materials written by experienced clinicians with answers produced by DeepSeek-V4. Independent experts assessed quality, accuracy, completeness and human sensitivity, while established tools measured readability and whether readers could act on the advice. The AI scored higher for overall quality and completeness and was comparable on several other measures.

The result suggests large language models could help clinicians prepare patient education, not replace professional judgement. DeepSeek performed relatively poorly on readability and humanistic care, and the study tested generated text rather than patients’ understanding, behaviour or health outcomes.

The researchers’ next step should be supervised trials with real patients, including checks for harmful errors and whether edited AI materials improve self-management.