Particle.news

Download on the App Store

Humans Adopt ChatGPT’s Preferred Vocabulary in Spoken Language, Preprint Finds

An arXiv preprint reveals that AI-preferred words have surged in online videos and podcasts over 18 months, suggesting ChatGPT patterns are reshaping everyday speech.

Image
As AI gets smarter, study says humans are starting to sound and talk more like ChatGPT
In 18 months, humans have begun using so-called “GPT words” that were previously not frequently used.
Image

Overview

  • Max Planck Institute researchers tracked the frequency of “GPT words” like delve, meticulous and realm in more than 360,000 YouTube videos and 771,000 podcast episodes since ChatGPT’s launch in 2022.
  • They defined “GPT words” by feeding emails, essays and academic papers into specific ChatGPT models and extracting the AI’s most favored terms.
  • The research highlights a bidirectional cultural feedback loop in which AI-driven language patterns are flowing back into human communication.
  • Authors warn that widespread adoption of AI-preferred vocabulary could accelerate linguistic homogenization and erode cultural diversity.
  • As a non-peer-reviewed preprint on arXiv, the study awaits validation and prompts calls for broader analysis of AI’s influence beyond academic speech sources.