Overview

• The AI tool increased immediate adenoma detection when active yet clinicians’ unassisted detection rates declined by roughly 20% compared with pre-AI baselines
• The observational study tracked outcomes at four Polish endoscopy centres over three months before and after AI deployment with some procedures randomly performed without the aid
• Nineteen highly experienced endoscopists, each having completed over 2,000 colonoscopies, took part, limiting the generalisability of findings to broader or less seasoned cohorts
• Study authors warn that reliance on AI likely prompted over-reliance, reducing clinicians’ motivation and focus during unaided procedures and risking deeper de-skilling as AI advances
• Experts urge preserved non-AI practice sessions, continuous performance monitoring and further randomized trials to balance AI benefits with skill retention