We’ve heard about upskilling and re-skilling due to AI — but how about de-skilling? A new study published this week found that doctors who frequently use AI to detect cancer in one medical procedure got significantly worse at doing so.

The researchers set out to discover whether continuous exposure to AI impacted doctors’ behavior when conducting colonoscopy, so they decided to assess “how endoscopists who regularly used AI performed colonoscopy when AI was not in use.”

The answer: Not so hot. The rate was about six percentage points lower.

The study was published in The Lancet Gastroenterology & Hepatology journal by medical professionals and researchers in countries including Poland, Norway, Sweden, the U.K., and Japan. It followed doctors at four endoscopy centers in Poland, which were part of a trial program focusing on AI’s use in colonoscopy for potential cancer prevention.

It raises questions about the use of AI in healthcare, when it helps and when it could hurt. Last week, The Verge reported on a Google healthcare AI model’s instance of potentially hallucinating a body part and where medical professionals think the industry will go from here.