Artificial intelligence may be making colonoscopies more accurate — but at a cost: doctors could be losing the very skills the technology was meant to support.

Routine use of AI-assisted colonoscopy systems can lead to a one-fifth drop in the ability of experienced endoscopists to detect adenomas — precancerous growths in the colon — without the technology, a new study in The Lancet Gastroenterology & Hepatology suggests.

Researchers examined colonoscopies of 1,443 patients who underwent non-AI assisted colonoscopy before and after the introduction of AI at four centers in Poland for six months. Before AI was rolled out, clinicians detected adenomas 28.4 percent of the time. Three months after AI became standard, the detection rate had fallen to 22.4 percent — a drop of six percentage points.

“Our results are concerning given the adoption of AI in medicine is rapidly spreading,” said study co-author Marcin Romańczyk from the Academy of Silesia in Poland. He urged additional research on the impact of AI on the skills of health professionals across different medical fields.

Colonoscopy is one of the most effective tools for preventing bowel cancer, allowing adenomas to be spotted and removed before they turn cancerous. AI assistance has generated excitement in recent years, with multiple trials showing higher detection rates when it’s used. But the study’s findings raise the concerns of “deskilling” — a slow erosion of expertise when clinicians rely too heavily on automated support. 

The findings also raise questions about earlier randomized controlled trials that reported higher adenoma detection rates with AI assistance than without it.

“It could be the case that non-AI assisted colonoscopy assessed in these trials is different from standard non-AI assisted colonoscopy as the endoscopists in the trials may have been negatively affected by continuous AI exposure,” said another co-author of the study, Yuichi Mori of the University of Oslo in Norway. 

The authors caution that the study’s observational design means other factors could have played a role.