Study Warns AI Diagnostic Tools May Undermine Clinician Skills

A recent study published in The Lancet Gastroenterology and Hepatology on August 12, 2025, indicates that regular use of artificial intelligence (AI) in medical diagnostics may lead to a decline in clinicians' independent diagnostic skills. The research, part of the AI in Colonoscopy for Cancer Prevention (ACCEPT) trial, was conducted across four endoscopy centers in Poland and involved 19 experienced endoscopists performing 2,177 colonoscopies between September 2021 and March 2022.

The study found that the adenoma detection rate (ADR)—a key quality indicator in colonoscopy—dropped from 28% to 22% in non-AI-assisted procedures after the introduction of AI tools. This suggests that clinicians may become over-reliant on AI assistance, potentially impairing their independent diagnostic abilities.

Marcin Romańczyk, a co-author of the study and assistant professor at the Medical University of Silesia, likened this phenomenon to the "Google Maps effect," where individuals lose navigational skills due to dependence on GPS technology. He explained that continuous exposure to AI assistance might lead to a similar decline in clinicians' diagnostic capabilities.

Omer Ahmad, a consultant gastroenterologist at University College Hospital London, emphasized the public health implications of the study's findings. He noted that even a 1% decline in ADR could significantly impact colorectal cancer prevention efforts. Ahmad called for caution in the real-world implementation of AI in medical diagnostics and advocated for more behavioral studies to understand how AI alters physician performance.

The integration of AI in healthcare has been both promising and challenging. While AI has demonstrated potential in improving diagnostic accuracy and efficiency, concerns about automation bias have emerged. Automation bias refers to the tendency of humans to favor suggestions from automated decision-making systems and to ignore contradictory information made without automation, even if it is correct. This bias can lead to diagnostic errors and a degradation of clinicians' diagnostic skills over time.

The ACCEPT trial's findings underscore the need for a cautious approach to AI integration in healthcare. While AI offers significant benefits, it is crucial to implement safeguards that prevent overreliance and ensure that clinicians maintain essential diagnostic skills. Ongoing research and adaptive training programs will be vital in achieving a harmonious balance between technological advancement and human expertise in medicine.