Post written by Michael B. Wallace, MD, MPH, from Mayo Clinic Florida, Jacksonville, Florida, USA.

This is part of a series of invited reviews focusing on the best articles of the past year. “Best of artificial intelligence in GI endoscopy” allows our endoscopic community to get a rapid overview of the top publications in the field of artificial intelligence (AI). This is particularly important because the volume and pace of articles is rapidly increasing to the point where it is nearly impossible for any individual to stay completely up to date on all aspects of AI and endoscopy.

Given the pace of innovation as well as the massive volume of articles in this field, we felt it would be very helpful for our community to have a succinct summary of some of the most important advances relevant to AI in endoscopy. On one hand, AI has matured to the point that we have multiple applications in our practice today but many more in various stages of the pipeline. We hear about tremendous applications to society in general and particularly in the health care field but want to know how we can use these in our practice to improve the care of our patients and the efficiency and quality of our work.

In this article, we summarize key advances in both established and emerging fields. Regarding established areas of AI, the most mature is polyp detection and to a lesser extent polyp classification systems. These systems are widely available now throughout the world and in use in many practices, including my own. We are now approaching almost 50 randomized controlled trials showing a nearly universal benefit for increased adenoma detection rate. Some initial real-world clinical studies suggest the benefit was limited, but other more recent real-world data continue to support that these technologies improve adenoma detection rate.

One of the limitations of the current technology is that it primarily increases detection of small, likely low-risk adenomas, although it does also increase detection of flat or nonpolypoid lesions, which can otherwise be very difficult to detect. There also has been some recent speculation based on 1 publication¹ that use of AI may lead to loss of independent skills, so-called “deskilling” by investigators performing traditional colonoscopy. This area remains open for investigation.

Other important advances in AI are workflow optimization including procedural documentation, office visit transcription, and quality metric automation. Many of the endoscopic AI systems are moving toward a platform-based technology where multiple applications can be embedded within the same platform.

We also hope to soon see other computer-aided detection systems for Barrett’s esophagus—associated neoplasia and pancreatobiliary endoscopy including classification of indeterminate bile duct strictures and pancreatic cystic neoplasms.

I would like to thank GIE Editor-in-Chief Dr Doug Adler for the opportunity to write this article. We hope it will provide insight for our Broad leadership.

Read the full article online.

The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.

1. Budzyń K, Romańczyk M, Kitala D, et al. Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study. Lancet Gastroenterol Hepatol 2025;10:896-903. ↩︎