Turns Out We're Starting To Talk Like ChatGPT Without Even Realising It. Check New Study
A study finds people are adopting ChatGPT’s tone and vocabulary in speech, pointing to a growing human-AI feedback loop that could reshape the way we communicate daily.
As AI becomes a fixture in daily life, it may be quietly reshaping how we talk. A recent study by the Max Planck Institute for Human Development suggests that people are beginning to imitate ChatGPT, both in vocabulary and tone. By analysing speech patterns in over a million YouTube and podcast transcripts, the researchers noticed a clear uptick in phrases commonly generated by large language models.
It’s not just about tech jargon; our natural way of expressing ideas might be slowly taking on an AI-like structure.
AI Vocabulary Creeping Into Real Conversations
The research, focused on terms the study dubs “GPT words,” such as delve, notably, and underscore. These words, once rare in casual conversation, have risen sharply in frequency since the public release of ChatGPT in late 2022. The researchers examined over 360,000 YouTube videos and 770,000 podcast episodes, comparing the language before and after ChatGPT’s launch.
Interestingly, this trend was visible not only in tech-focused content but across a wide variety of genres, from lifestyle to entertainment. According to lead author Levin Brinkmann, “Delve is only the tip of the iceberg.”
A Feedback Loop Between Humans and Machines
The study refers to this shift as a “closed cultural feedback loop.” Humans influence AI models by training them on public text and speech. But now, the influence appears to go both ways. As people consume more AI-generated content, directly or indirectly, they start adopting its sentence structures, pacing and phrasing.
This loop creates a linguistic mirror where machine-learned styles become part of human communication, even when the AI isn’t being used in the moment.
Why This Matters for Language Diversity
While AI-style communication may sound clearer and more formal, some experts fear that it could flatten linguistic diversity. Regional idioms, emotional expressions and spontaneous quirks might fade, replaced by a more uniform, sterile mode of speaking.
As the study’s co-authors argue, we may need to rethink how we interact with AI, not just in what it tells us, but in how it changes us.
Top Headlines
