The Impact of Large Language Models on Academic Writing

Large language models (LLMs) are increasingly used in academia for tasks such as drafting, editing, and summarizing. While these tools can improve productivity and accelerate the research and writing processes, their repercussions on academic writing conventions are a subject that requires further investigation. This work examines the impact of LLMs on academic writing by analyzing text similarity and linguistic trends in scientific research from 2020 to 2024, using several metrics to assess over 60,000 abstracts. Findings indicate that while LLMs have not yet led to a significant homogenization of academic abstracts in terms of lexicon and overall meaning, they are reshaping writing styles in distinct ways. Through the study of AI-generated revisions of existing abstracts, linguistic features associated with the “LLM writing style” were characterized, including a preference for conciseness, increased lexical diversity and complexity, and a more direct, immediate tone, often at the cost of readability and cohesion. Temporal analyses confirm most of these trends, highlighting the evolving role of LLMs in shaping the academic discourse, and raising important questions about originality, accessibility and the future of scientific research.