Abstract
Biomedical review articles are important for education, evidence synthesis, and clinical decision-making. Large language models such as ChatGPT may contribute to a surge in the number of AI-written reviews, raising concerns about the quality and reliability of scientific content. We searched PubMed for open-access reviews in inflammatory bowel disease (IBD) published before ChatGPT (period 1), during the first year after ChatGPT’s release (period 2, 2023–2024), and the year after (period 3, 2024–2025), and randomly selected 60 articles for analysis from each period. Using an AI text detector, no articles were identified as AI-written in period 1, versus 1.7% and 6.7% in periods 2 and 3, respectively. None of the potentially AI-written articles disclosed AI use. None of them had an IBD expert as the first author; only one article from period 2 and one from period 3 had an IBD expert among the authors. While the absolute number of AI-written reviews remains relatively low at present, the growing trend is a cause for concern regarding the potential decline of the quality and trustworthiness of biomedical review articles. Safeguards should be established to maintain the values of review articles.
Keywords
Get full access to this article
View all access options for this article.