Gemini and ChatGPT outperform Claude in matching expert consensus on prostate cancer guideline questions
Study · UroToday · 2026-11-09. Number 2 in the UrologyWatch Top 5 for November 9, 2026.
In a two-round Delphi study, 11 experts evaluated 25 guideline-based prostate cancer questions answered by three LLMs: consensus agreement was reached for 23/25 (92%) Gemini responses, 20/25 (80%) ChatGPT responses, and 14/25 (56%) Claude responses (p=0.0033), with explicit disagreement noted for 3 Claude answers after round two.
Why it matters: Suggests LLMs like Gemini and ChatGPT may assist in interpreting routine guideline content but require expert oversight, with notably lower reliability from Claude, especially in complex scenarios involving advanced imaging or genetic testing.
Limitation: The study relied on a modified Delphi consensus process among 11 experts rather than validated clinical outcomes, and performance varied with guideline complexity.
Source: UroToday · PubMed 42980497 · DOI 10.57264/cer-2026-0081
Loading today's edition…