Miller School of Medicine Gastroenterologists Study Use of ChatGPT for Surveillance Colonoscopy Intervals (Miller School of Medicine)

University of Miami researchers report that a customized large language model achieved ~95% accuracy in recommending guideline-based follow-up colonoscopy intervals — matching GI physicians in most cases. The tool, validated in 1,000 patient colonoscopy datasets and published in AJG, could reduce errors and save clinicians time, though experts stress a human-in-the-loop remains essential as hallucination and complex cases still challenge current models.