Real-World Concordance of the European Association of Urology Guidelines Bot With Multidisciplinary Decision-Making in Prostate Cancer: A Prospective Study

Abstract Background Guideline-based artificial intelligence tools may support oncological decision-making, but their performance in real-world multidisciplinary care remains uncertain. We evaluated the concordance between recommendations generated by the European Association of Urology (EAU) Guidelines Bot and Findings: We prospectively evaluated 156 patients with prostate cancer discussed at a multidisciplinary tumor board between June 2025 and April 2026 using a standardized clinical prompt. Recommendations were classified as complete concordance (level 1), correct but incomplete (level 2), partially correct with erroneous recommendations (level 3), or completely incorrect (level 4). Complete concordance was observed in 74.3% of cases, while 89.7% were considered clinically acceptable (levels 1–2). Overall, 39 cases showed some degree of discrepancy (levels 2–4). The main sources of discrepancy were lack of awareness of local clinical trials or protocols (20.5%), inadequate adaptation to frailty or severe comorbidity (15.4%), omission of genetic testing (10.3%), and omission of required diagnostic or restaging steps (10.3%). Conclusions The EAU Guidelines Bot showed high concordance with multidisciplinary decision-making, but limitations emerged when recommendations required patient-specific or locally contextualized information. These findings support its use as a decision-support tool rather than a substitute for multidisciplinary clinical judgment.