PURPOSE: The integration of artificial intelligence (AI) in scientific writing has rapidly evolved, raising concerns about the prevalence of AI-generated content in peer-reviewed medical literature (1). Some journals have even published manuscripts that were entirely AI-generated in order to demonstrate its capabilities to its audiences (2). The field of plastic and reconstructive surgery relies heavily on high-quality research to guide clinical practice. However, the increasing accessibility of AI writing tools introduces questions about the authenticity, accuracy, and ethical implications of AI-assisted manuscript generation. While AI can enhance efficiency and language clarity, excessive reliance may compromise scientific integrity, introduce biases, or obscure the true intellectual contributions of authors. This study aims to quantify the presence of AI-generated text in the leading journal within our field, Plastic & Reconstructive Surgery (PRS), providing insight into the extent of AI's role in academic publishing. Understanding these trends is essential for maintaining rigorous scholarly standards, ensuring transparency, and guiding future editorial policies to uphold the integrity of evidence-based surgical research. Further, understanding the ability of AI-text detection software may prove useful to editors wishing to screen manuscript for AI-content prior to publication. METHODS: The authors of this study randomly selected 6 issues published within PRS in 2014 and 2024. Articles that fell into one of nine categories were included for further analysis while other article types were excluded: original research article, systematic review/meta-analysis, reviews, ideas and innovations, special topics, viewpoints, letters/replies, editorials, discussions. Manuscripts' text was analyzed in the AI-text detection software ZeroGPT in order to determine the prevalence of AI-generated material in each article (3). Research articles and systematic reviews/meta-analyses were further analyzed by manuscript section (Abstract, Background, Methods, Results, Discussion, Conclusion). Articles published in 2014 served as the study's control as AI's capacity to create generative human-like text had not yet been developed. A two-tailed student's T-test was used to test for significance. RESULTS: A total of 273 articles were identified in 2014 and 205 in 2024 for inclusion. The mean percent of reported AI-generated content across all article types was 20.8% (SD = 22.8) in 2014 and 15.2% (SD = 14.5) in 2024. A significant difference was observed between these two values (p = 0.002), indicating a significant decrease AI-generated content in the 2024 group. On average, article abstracts were flagged as having the greatest amount of AI-content (37.9%), followed by the results section (36.7%). CONCLUSION: The results of this study were unexpected as there was a significant decrease in detectable AI material in manuscripts published in 2024. We expected there to be at least a modest increase in detected AI-content in the 2024 group. The results of this study indicate that the use of AI-generated content in PRS manuscripts has not yet occurred. Alternatively, the results of this study may indicate further work is needed in AI-detection software before it can be reliably utilized by journal editors to screen manuscripts for AI-generated content.
Bohler et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: