The Silent Author: A Pilot Study Detecting AI-Assisted Writing in Plastic and Reconstructive Surgery Journals

Area of Science:

• Medical publishing
• Artificial intelligence in medicine
• Plastic and reconstructive surgery

Background:

• Generative artificial intelligence (AI) adoption in academic publishing raises concerns for surgical decision-making, research credibility, and patient safety.
• No prior pilot studies have assessed AI's presence in plastic and reconstructive surgery literature.
• Current AI detection tools are imperfect, and journal policies on AI disclosure lack consensus, creating a gap in oversight.

Purpose of the Study:

• To pilot an analysis of AI-like textual characteristics in plastic and reconstructive surgery publications.
• To compare the prevalence of AI-like text in contemporary publications versus a pre-AI control cohort.

Main Methods:

• A pilot analysis sampled 10% (n=67) of articles from leading plastic surgery and burn journals (July 2024-July 2025).
• A matched control cohort from 2014-2015 was assembled using identical criteria.
• Articles were analyzed using a RoBERTa classifier and perplexity evaluation to identify potential AI-generated text.

Main Results:

• 18 of 67 contemporary articles (26.9%) and 20 subsections (5.3%) contained AI-like textual characteristics.
• Flagged content was most prevalent in methods (9.1%) and abstracts (7.5%).
• A statistically significant ~6-fold increase in AI-like text was observed compared to the 2014-2015 control cohort (4.5% of papers, 0.9% of subsections).

Conclusions:

• AI-like textual characteristics are significantly more frequent in current plastic and reconstructive surgery publications than in a pre-AI era cohort.
• These findings emphasize the critical need for enhanced journal disclosure policies, standardized reporting guidelines, and reviewer education.
• Responsible integration of AI into surgical publishing requires proactive measures to maintain research integrity and patient safety.