ai
Most biomedical publications show signs of LLM-assisted writing
- Source
- arXiv — Computers and Society
- Published
- Last verified
- 12 Aug 2026
- Confidence
- Moderate
- Evidence
- Original document retained
- Reading time
- 1 min
- Country
- International
- Relevant to
- Research & Evidence, Policy & Regulation, Risk & Compliance, Technology & Data
Executive summary
What happened, and why should leadership care?
A research paper from arXiv suggests that a significant majority of biomedical publications now exhibit characteristics consistent with Large Language Model (LLM)-assisted writing. The study, utilizing a novel method based on changing word frequencies, estimates that by late 2025, approximately 89% of open-access biomedical papers from Pubmed Central will show signs of LLM influence. This trend, while potentially beneficial for overcoming language barriers, also raises concerns regarding research integrity and potential misconduct.
Why this matters
Why is this strategically important?
The widespread adoption of LLM-assisted writing in scholarly publications, particularly in critical fields like biomedicine, fundamentally impacts the perceived integrity and reliability of research outputs. This necessitates a strategic re-evaluation of publication standards, ethical guidelines, and verification processes across academic and research institutions to maintain trust and scientific rigor.
Key insights
What should be noted from the evidence?
- LLM-powered chatbots and agents are widely adopted for academic writing.
- LLM assistance can aid in overcoming language barriers but also poses risks of misconduct and fraud.
- A new unbiased methodology was developed to estimate LLM usage in textual corpora based on changing word frequencies.
- Application of this method to open-access biomedical papers from Pubmed Central indicates an estimated 89% prevalence of LLM-assisted writing by the end of 2025.
Evidence and confidence
How far can this assessment be trusted?
Moderate confidence. Provenance established; supporting evidence remains partial.
Analysis is prepared editorially by Aziz Shuaib Ausi. The original publication remains the authoritative record, and executive judgement remains entirely human.
Source
Where does this originate?
Reported by arXiv — Computers and Society · International. This briefing summarises the publication for executive use; the document itself is not reproduced here.
Read the original publication