(493) Evaluating GPT4-0's Responses to Andrology Patient Questions in Outpatient Clinic Compared to European Association of Urology Guidelines

Introduction Generative artificial intelligence (AI) models such as GPT-4.0 are increasingly used by patients to obtain healthcare information, though their accuracy remains uncertain. Objective This study compared GPT-4.0 responses to patient questions from an andrology outpatient clinic with the 2025 European Association of Urology (EAU) guidelines. Methods Prospective face-to-face and telephone andrology clinics were conducted between March and May 2025. Patients were asked to list questions they would submit to GPT-4.0 about their andrology condition. All questions were grouped by condition and theme, then entered into GPT-4.0. The quality of answers and compliance with the EAU Sexual and Reproductive Health 2025 guidelines were assessed. Each response was classified as correct, missing key elements, or incorrect/misleading. Results A total of 208 patients were seen. After removing duplicates, 15 unique questions remained, covering erectile dysfunction (ED), Peyronie’s disease (PD), and male factor infertility (MFI). GPT-4.0 produced detailed answers on the causes, investigations, and treatments for these conditions. However, 7 responses (46.7%) lacked key elements, and 10 (66.7%) contained incorrect or misleading information compared with EAU guidance. The study is limited by its small sample size and the use of a single AI chatbot. Conclusions GPT-4.0 can generate comprehensive responses to patient questions, but accuracy varies. Interpretation depends on user understanding, and some outputs diverge from established pharmacological and pathophysiological guidance. AI-generated health information should therefore be interpreted cautiously and verified against clinical guidelines. Disclosure No