AI writing assistants shrink linguistic diversity and blur personal identity | Nature Human Behaviour

Large language models (LLMs) used as writing assistants homogenize writing style. Studies across 880,000 texts show that, following the release of ChatGPT, writing became less varied as artificial intelligence (AI) use spread. The LLMs preserved meaning but compressed the variety of language and biased author traits inferred from the words. This is a preview of subscription content, access via your institution Access options Access through your institution Access Nature and 54 other Nature Portfolio journals Get Nature+, our best-value online-access subscription 27,99 € / 30 days cancel any time Learn more Subscribe to this journal Receive 12 digital issues and online access to articles 111,21 € per year only 9,27 € per issue Learn more Buy this article Purchase on SpringerLink Instant access to the full article PDF. 39,95 € Prices may be subject to local taxes which are calculated during checkout Fig. 1: As use of LLMs increases, writing becomes less varied and the identity it carries shifts. Subjects Human behaviour Interdisciplinary studies Science, technology and society References Kennedy, B. et al. Moral concerns are differentially observable in language. Cognition 212 , 104696 (2021). This paper shows that personal characteristics such as moral concerns leave measurable, systematic traces in everyday word use. Article PubMed Google Scholar Pennebaker, J. W. The Secret Life of Pronouns: What Our Words Say about Us (Bloomsbury, 2013). This book gives a foundational account of how the small, automatic words we use reveal personality, social background and psychological state. Hall, M. et al. Bias amplification in image classification. In Proc. Workshop on Trustworthy and Socially Responsible Machine Learning, NeurIPS 2022 https://openreview.net/forum?id=lwG9sG4sbVC (2022). This paper shows systematically that machine-learning models amplify the biases present in their training data. Wang, A., Morgenstern, J. & Dickerson, J. P. Large language models that replace human participants can harmfully misportray and flatten identity groups. Nat. Mach. Intell. 7 , 400–411 (2025). This paper documents how LLMs flatten and misrepresent identity groups, and complements our evidence that LLM rewriting biases the personal traits inferred from text. Article Google Scholar Hans, A. et al. Spotting LLMs with binoculars: zero-shot detection of machine-generated text. Proc. Mach. Learn. Res. 235 , 17519–17537 (2024). The paper introduces the zero-shot detector we used to estimate the prevalence of AI-generated text at scale, which enabled our real-world trend analysis. Google Scholar Download references Additional information Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This is a summary of: Sourati, Z. et al. The shrinking landscape of linguistic diversity in the age of large language models. Nat. Hum. Behav . https://doi.org/10.1038/s41562-026-02550-0 (2026). Rights and permissions Reprints and permissions About this article Cite this article AI writing assistants shrink linguistic diversity and blur personal identity. Nat Hum Behav (2026). https://doi.org/10.1038/s41562-026-02549-7 Download citation Published : 24 August 2026 Version of record : 24 August 2026 DOI : https://doi.org/10.1038/s41562-026-02549-7 Share this article Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy shareable link to clipboard Provided by the Springer Nature SharedIt content-sharing initiative