Examining the frequency of artificial intelligence generated content in anesthesiology and intensive care journal publications: A cross sectional study.
Journal:
Medicine
PMID:
39993120
Abstract
The emergence of artificial intelligence (AI)-based linguistic models has revolutionized academic writing, prompting concerns about integrity. In response, AI-powered text authenticity detectors have been developed. This study examines AI tool usage in anesthesiology and intensive care journals. 1268 articles from 86 journals in "Anesthesiology" and "Anesthesiology and Intensive Care" were analyzed using Copyleaks and ZeroGPT. English abstracts published between April 18 and May 18, 2023, were scrutinized. ZeroGPT and Copyleaks found average AI usage at 25.1% ± 27.5 and 10.5% ± 15.9, respectively. 16.8% of articles were "human-written," while 83.2% were "AI-assisted". AI assistance correlated positively with abstract length and was more common among nonnative English speakers (P < .001). It was also prevalent in high-impact and science citation index-indexed journals (P < .01; P < .001). This study underscores the widespread adoption of AI tools in academic writing, particularly among nonnative English authors and in high-impact journals, emphasizing the need for improved detection mechanisms and regulatory guidelines.