Study Finds 13% of 2024 Biomedical Abstracts Likely Used AI Assistance

At least 13 percent of biomedical research abstracts published in 2024 may have been influenced by artificial intelligence, particularly large language models (LLMs), according to a new study by researchers at the University of Tübingen, Germany.

The study, which analyzed over 15 million biomedical research papers from 2010 to 2024, found a significant shift in writing style, especially in the choice of vocabulary. Researchers identified a rise in the use of so-called “style words”—terms and phrasing patterns frequently generated by LLMs like ChatGPT and other AI-based tools.

The study suggests that such linguistic markers are now common enough to estimate that more than one in ten abstracts published this year were likely aided by AI.

“Large language models have caused a drastic shift in the vocabulary used in academic writing,” the researchers noted, adding that speculation about AI’s growing role in scientific literature has now been supported with measurable data.

Trained on massive datasets, language models can generate, summarize, or edit scientific text based on natural language prompts. While they offer productivity and efficiency, the rise of AI-generated content has sparked concerns around academic authenticity, plagiarism, and scientific accuracy.

This is one of the first large-scale quantitative studies to examine the stylistic fingerprints of AI in academic publishing. While the presence of AI-generated content doesn’t imply reduced quality, it raises questions about transparency, authorship norms, and ethical standards in research communication.

The findings have reignited the debate about whether scientific journals should mandate the disclosure of AI assistance in the writing process, as the boundary between human and machine authorship continues to blur.