Potential Risks of Routine AI Assistance on Endoscopist Skills in Colonoscopy

The integration of artificial intelligence (AI) in colonoscopy practices has been associated with a concerning decline in the diagnostic skills of endoscopists, particularly in identifying precancerous lesions or adenomas. A recent study published in The Lancet Gastroenterology & Hepatology highlights that regular use of AI during colonoscopies may diminish the ability of experienced endoscopists to detect adenomas without AI support.

Colonoscopy is a vital tool in preventing bowel cancer by enabling the detection and removal of adenomas. While numerous trials have demonstrated that AI assistance significantly boosts adenoma detection rates—leading to enthusiasm for its implementation—the long-term impact on clinicians’ skills remains underexplored. This study is among the first to suggest that continual AI assistance could inadvertently weaken clinicians’ ability to perform accurate, independent assessments.

Conducted across four Polish colonoscopy centers between September 2021 and March 2022, the research compared colonoscopies performed before and after the adoption of routine AI assistance. A total of 1,443 procedures were analyzed, with 795 conducted prior to AI introduction and 648 afterward. The procedures involved highly experienced endoscopists, each with over 2,000 performed colonoscopies.

Results indicated a significant decrease in adenoma detection rates: from 28.4% before AI use to 22.4% after, representing a 20% relative reduction. In contrast, colonoscopies with AI assistance achieved a detection rate of 25.3%. Experts like Professor Yuichi Mori emphasized that these findings challenge previous controlled trials showing higher detection rates with AI, suggesting the possibility that continuous AI exposure may negatively influence clinician performance.

The study authors warn that this phenomenon of 'deskilling' could have serious implications for patient outcomes. They strongly advocate for further research across various medical fields to understand the dynamics between AI tools and clinician skills. It’s crucial to identify contributing factors and develop strategies to mitigate potential adverse effects.

While this study focused on experienced endoscopists, its results may have broader implications, especially for less experienced professionals. As AI becomes more widespread, safeguarding fundamental clinical skills becomes imperative to ensure high-quality patient care.

Commenting on these findings, Dr. Omer Ahmad from University College London stressed the need for cautious integration of AI in medicine, highlighting that while AI can enhance outcomes, it might also lead to the gradual erosion of essential skills if not managed carefully.

This research underscores an important aspect of AI adoption: balancing technological benefits with maintaining clinician competence, emphasizing the need for ongoing training and evaluation in medical practice.

For more details, the full study can be accessed through The Lancet Gastroenterology & Hepatology: [DOI: 10.1016/S2468-1253(25)00133-5]. Source: https://medicalxpress.com/news/2025-08-routine-ai-loss-skills-health.html