In a multicentre observational study across four Polish endoscopy centres participating in the ACCEPT trial, Krzysztof Budzyń, Marcin Romańczyk, Diana Kitala, Paweł Kołodziej, Marek Bugajski, and Hans O. Adami examined whether continuous exposure to AI-assisted colonoscopy affects endoscopist performance when AI is not in use. The team analyzed 1,443 non-AI-assisted diagnostic colonoscopies—795 performed before and 648 after AI introduction.
Results showed a drop in adenoma detection rate (ADR) from 28.4% pre-AI to 22.4% post-AI exposure (absolute difference –6.0%, p=0.0089). Multivariable analysis identified AI exposure (OR 0.69), male patient sex, and age ≥60 years as independent factors influencing ADR.
The authors concluded that prolonged AI exposure may lead to endoscopist deskilling in standard colonoscopy, raising concerns about overreliance on AI tools. The study was funded by the European Commission and the Japan Society for the Promotion of Science.