Does GenAI Rewrite How We Write? An Empirical Study on Two-Million Preprints

TL;DR

The paper investigates whether Generative AI rewrites how we write by analyzing over 2.1 million preprints across four major repositories. It employs an integrated framework combining interrupted time-series analysis, collaboration metrics, linguistic profiling, and topic modeling to track changes in volume, style, and topics. Key findings show faster submission and revision cycles, modest increases in linguistic complexity of abstracts, and a surge of AI-related topics, with the strongest effects in computationally heavy fields. The results suggest GenAI acts as a selective catalyst that reinforces existing strengths and widens disciplinary divides, underscoring the need for governance to preserve trust, fairness, and accountability.

Abstract

Preprint repositories become central infrastructures for scholarly communication. Their expansion transforms how research is circulated and evaluated before journal publication. Generative large language models (LLMs) introduce a further potential disruption by altering how manuscripts are written. While speculation abounds, systematic evidence of whether and how LLMs reshape scientific publishing remains limited. This paper addresses the gap through a large-scale analysis of more than 2.1 million preprints spanning 2016--2025 (115 months) across four major repositories (i.e., arXiv, bioRxiv, medRxiv, SocArXiv). We introduce a multi-level analytical framework that integrates interrupted time-series models, collaboration and productivity metrics, linguistic profiling, and topic modeling to assess changes in volume, authorship, style, and disciplinary orientation. Our findings reveal that LLMs have accelerated submission and revision cycles, modestly increased linguistic complexity, and disproportionately expanded AI-related topics, while computationally intensive fields benefit more than others. These results show that LLMs act less as universal disruptors than as selective catalysts, amplifying existing strengths and widening disciplinary divides. By documenting these dynamics, the paper provides the first empirical foundation for evaluating the influence of generative AI on academic publishing and highlights the need for governance frameworks that preserve trust, fairness, and accountability in an AI-enabled research ecosystem.

Figures (14)

• Figure 1: Overview of research questions
• Figure 2: LLM transformer architecture
• Figure 3: Methodology
• Figure 4: Composition of the 2016–2025 corpus across repositories. The dataset integrates four major preprint platforms: arXiv, bioRxiv, medRxiv, and SocArXiv.
• Figure 5: Submission across repositories. The plots summarize monthly submission counts with a 3-month moving average, cumulative totals, and year-over-year growth rates. Vertical markers denote the GPT-3 release (June 2020) and the ChatGPT release (November 2022).