Quality, readability, and potential safety risks of epididymitis information generated by six artificial intelligence chatbots: a cross-sectional comparative study

Background AI chatbots are increasingly used for medical information, but the quality, readability, and safety of disease-specific responses remain uncertain. We compared epididymitis information generated by six widely accessible AI chatbots. Methods Nine of 25 Google Trends queries identified worldwide from August 2021 to August 2026 were each submitted once to ChatGPT, Gemini, Perplexity, Copilot, Kimi, and DeepSeek. Responses were assessed with DISCERN, EQIP, JAMA, GQS, and six readability indices. Between-model differences in DISCERN, EQIP, and the readability indices were assessed using Friedman tests; significant outcomes underwent paired Wilcoxon signed-rank tests, with Holm correction for multiple comparisons. JAMA and GQS were summarized descriptively. Potentially harmful misinformation was assessed exploratorily. Results DISCERN, EQIP, and all six readability indices showed significant overall model effects after Holm correction, but no post hoc pairwise comparison remained statistically significant. Copilot had the highest median EQIP score (70.00). JAMA scores were 0 in 52/54 responses (96.3%), and none of the 54 responses met sixth-grade readability benchmarks. Potentially harmful information occurred in 4/54 responses (7.4%), all involving antimicrobial recommendations. Conclusion In this exploratory cross-sectional snapshot, the chatbots differed in selected quality measures but shared poor source transparency and excessive reading complexity. They may supplement general patient information but should not replace professional assessment or guideline-based care.