Related Experiment Video
Updated: Apr 30, 2026
Flexible Colonoscopy in Mice to Evaluate the Severity of Colitis and Colorectal Tumors Using a Validated Endoscopic Scoring System
Published on: October 16, 2013
Precision of Chatbot Generative Pretrained Transformer Version 4-Generated References for Colon and Rectal Surgical
Aaron L Albuck1, Chad M Becnel2, Daniel J Sirna1
1School of Medicine, Tulane University, New Orleans, Louisiana.
Introduction:
The objective is to assess the precision of references generated by Chatbot Generative Pretrained Transformer version 4 (ChatGPT-4) in scientific literature pertaining to colon and rectal surgery.
Methods:
Ten frequently studied keywords pertaining to colon and rectal surgery were chosen: colon cancer, rectal cancer, anal cancer, total neoadjuvant therapy, diverticulitis, low anterior resection, transanal minimally invasive surgery, ileal pouch anal anastomosis, abdominoperineal resection, and hemorrhoidectomy. ChatGPT-4 was prompted to search for the most representative citations for all keywords. After this, two separate evaluators meticulously examined the outcomes each key element, awarding full accuracy to generated citations in which there was no discrepancies in any of the fields when cross-referenced with the Scopus, Google, and PubMed databases. References from ChatGPT-4 underwent a thorough review process, which involved careful examination of key elements such as the article title, authors, journal name, publication year, and Digital Object Identifier (DOI).
Results:
Forty-one of the 100 references generated by were fully accurate; however, but none included a DOI. Partial accuracy was observed in 67 of the references, which were identifiable by title and journal. Performance varied across specific keywords; for example, references for colon and rectal cancer were 100% identifiable by title and journal, but no term had 100% accuracy across all categories. Notably, none of the generated references correctly listed all authors. Conducted within a short timeframe during which ChatGPT4 is rapidly evolving and updating its knowledge base.
Conclusions:
While ChatGPT-4 offers improvements over its predecessors and shows potential for use in academic literature, its inconsistent performance across categories, lack of DOIs, and irregularities in authorship listings raise concerns about its readiness for application in the field of colon and rectal surgery research.
