Researchers at Berlin’s Max Planck Institute for Human Development have unveiled compelling evidence that the release of ChatGPT has had a measurable impact on the way people speak in academic settings. By transcribing and analyzing roughly 280,000 English-language videos from more than 20,000 YouTube channels affiliated with academic institutions, the team identified a pronounced increase—up to 51 percent—in the use of certain English words closely associated with AI language models. Words like meticulous, delve, realm, and adept now appear far more frequently in presentations, lectures, and talks, suggesting that humans are unknowingly adopting vocabulary and phrasing patterns popularized by large language models (LLMs).
This study marks the first empirical demonstration that AI-generated language is influencing human spoken communication on a large scale. The researchers warn that this trend could have far-reaching consequences. On one hand, it may streamline and elevate formal communication; on the other, it could unintentionally erode linguistic diversity, reducing the richness and variety of human expression. Moreover, the integration of AI language into everyday speech raises concerns about how such technologies might be exploited for manipulation, propaganda, or social engineering. The authors emphasize the need for urgent, in-depth exploration of the feedback mechanisms between AI behavior and human cultural evolution. Lead investigator Hiromu Yakura noted that speakers generally remain unaware of this subtle shift in their speech patterns, calling it a process whereby “virtual vocabulary” from AI systems is being absorbed and normalized within daily language, fundamentally altering how we communicate without conscious realization.