Readability, information quality, and safety of AI chatbot responses to common questions about ostomy care: A cross-sectional digital health evaluation
Yu Zhang, Xia Li, Jiayi Su, LI Yuan, Aixin Zhou, Liuyin Chen, Hua ZhangBackground
Patients with ostomies often require timely, understandable, and reliable guidance for self-care after surgery. Although AI chatbots are increasingly used as accessible sources of health information, the quality and safety of their responses for ostomy care remain uncertain.
Objective
To compare information quality, readability, and text-level safety across five chatbots answering common ostomy-care questions.
Methods
We submitted forty-six matched patient-facing ostomy-care questions once to five chatbots (ChatGPT, Gemini, Perplexity AI, DeepSeek, and Doubao) through official web interfaces on March 31, 2026, yielding 230 first responses. Responses were assessed with DISCERN, EQIP, Global Quality Score (GQS), JAMA benchmarks, six readability indices, and a guideline-based audit. Citation review was limited to Perplexity AI, the only system displaying source links. Comparisons used Friedman, Wilcoxon, and Cochran’s Q tests with multiplicity adjustment.
Results
Scores differed across all four measures of quality and all six readability indices after Holm correction (adjusted
Conclusions
None was consistently suitable as a stand-alone source of ostomy education. Their responses may support patient education after professional review but should not replace clinician-guided instruction.