AI chatbots for patient education in localized prostate cancer radiotherapy: A comparative quality and readability analysis.

OBJECTIVES: This study aimed to conduct a comparative quality assessment of information provided by widely used artificial intelligence chatbots (AICs) regarding radiotherapy for localized prostate cancer, with a focus on reliability, readability, and patient-centeredness. METHODS: Five publicly accessible AICs (ChatGPT, Perplexity, Gemini, DeepSeek, and Copilot) were evaluated using three standardized questions addressing treatment effectiveness, administration, and side effects of radiotherapy. All responses were generated during a predefined study period using publicly available chatbot versions. The outputs were anonymized and independently assessed by five radiation oncologists using validated instruments, including DISCERN, the Patient Education Materials Assessment Tool for Printed Materials (PEMAT-P), the Web Resource Rating (WRR) scale, the Coleman-Liau Index, and a guideline consistency Likert scale. This design allowed for a structured, comparative evaluation of chatbot-generated educational content. RESULTS: Overall, Gemini and Perplexity consistently demonstrated higher information quality and reliability, while ChatGPT, DeepSeek, and Copilot showed moderate performance. Although several chatbots achieved acceptable understandability scores, all responses required college-level literacy, limiting accessibility for many patients. Differences across chatbots reflected variability in transparency, reference citation, and actionable guidance rather than factual accuracy alone. CONCLUSIONS: AI chatbots may serve as supplementary sources of information for patients with localized prostate cancer; however, their current outputs are not suitable for unsupervised patient education in radiotherapy settings. The lack of stage-specific recommendations and high literacy demands pose potential risks, including misunderstanding of treatment options and unrealistic expectations. Clinicians should be cautious when patients rely on chatbot-generated information, and developers should prioritize patient-centered language, stage-specific content, and guideline-aligned updates. These findings highlight the need for professional oversight before integrating AI chatbots into routine radiotherapy patient education.