Routine reliance on artificial-intelligence software during colonoscopies may erode clinicians’ core diagnostic skills, according to research published in The Lancet Gastroenterology & Hepatology. The study tracked 19 experienced endoscopists at four centres in Poland who alternated between using and not using an AI tool that highlights potential precancerous polyps. Across roughly 2,200 procedures, doctors detected adenomas in 28 % of examinations during the three months before the algorithm was introduced. After three months of regular AI use, the rate fell to 22 % in cases where the software was switched off—a 20 % relative decline and a drop of six percentage points compared with the pre-AI baseline. Researchers described the finding as the first clinical evidence of a “deskilling” effect from medical AI and warned that over-reliance could jeopardise patient outcomes if systems fail or become unavailable. Lead author Marcin Romańczyk said further work is needed “to understand and prevent skill loss as AI spreads across medical specialties.”