This research quantifies the impact of ChatGPT on academic writing by analyzing vocabulary patterns in over 14 million PubMed abstracts.

Key findings:

• Scientists are increasingly using ChatGPT for academic writing, detectable through specific vocabulary patterns
• LLM-generated text contains distinctive linguistic markers that differ from traditional human writing
• This trend raises important questions about authorship, authenticity, and academic integrity
• The methodology provides an unbiased approach to identifying AI-assisted content at scale

For educators and academic institutions, this research offers critical insights into evolving writing practices and highlights the need for thoughtful policies on AI use in scholarly contexts.

Delving into ChatGPT usage in academic writing through excess vocabulary