Comparative performance of artificial intelligence chatbots in patient education for robot-assisted radical...

Patients undergoing robot-assisted radical prostatectomy require complex counseling regarding treatment selection, cancer control, urinary continence, sexual function, postoperative therapy, and long-term follow-up. The suitability of artificial intelligence chatbots for providing such information remains uncertain. Twenty clinician-developed patient-education questions on robot-assisted radical prostatectomy were submitted to ChatGPT-o3, DeepSeek-V4, Claude Sonnet 5, and Gemini 3.5 Pro. Responses were evaluated using DISCERN, the Ensuring Quality Information for Patients (EQIP) tool, the Global Quality Scale (GQS), and the Journal of the American Medical Association (JAMA) benchmark criteria. Readability was assessed using six established indices. Between-model differences were analyzed using nonparametric tests with corrected post hoc comparisons. Information-quality scores differed significantly across models for DISCERN, EQIP, and GQS. DeepSeek-V4 achieved the highest DISCERN (61.90 ± 2.92), EQIP (91.75 ± 8.32), GQS (4.50 ± 0.51), and JAMA (3.00 ± 0.00) scores. Gemini 3.5 Pro showed relatively lower information-quality scores among the evaluated models. ChatGPT-o3 and DeepSeek-V4 demonstrated relatively favorable readability on different indices; however, all models exceeded the recommended sixth-grade reading level, and all Flesch Reading Ease scores remained below the easy-to-read threshold. AI chatbots varied substantially in the quality, transparency, and readability of information on robot-assisted radical prostatectomy. Although some models provided comparatively stronger patient education, none consistently produced sufficiently accessible information. These tools may supplement, but should not replace, individualized counseling by urologists and multidisciplinary prostate cancer teams.