Particle.news
Download on the App Store
4 ARTICLES
·
4h ago

AI-Favored Words Increase in Human Speech, Researchers Report

Released on arXiv without peer review, the research highlights AI’s expanding influence on human vocabulary.

Image
In 18 months, humans have begun using so-called “GPT words” that were previously not frequently used.
Image

Overview

  • Researchers identified “GPT words” by having ChatGPT polish millions of emails, essays and articles and selecting terms such as “delve,” “realm” and “meticulous” that the model favored.
  • They analyzed over 360,000 YouTube talks and 771,000 podcast episodes recorded before and after ChatGPT’s November 2022 debut to track shifts in word usage.
  • The preprint, posted to arXiv in July 2025, is not yet peer reviewed and focuses on specific GPT models, which may limit its broader applicability.
  • Study authors warn that humans may imitate AI language patterns when they perceive the technology as authoritative, creating a feedback loop between machine outputs and spoken communication.
  • The team cautions that continued convergence on AI-preferred vocabulary risks accelerating linguistic homogenization and eroding cultural diversity.