How ChatGPT Is Quietly Changing the Way We Speak
When ChatGPT was released in late 2022, it made headlines for reaching 100 million users in just two months, a milestone that earned it the title of the fastest-growing consumer application in history. Most of the attention at the time focused on its utility: students using it to write essays, professionals drafting emails, coders speeding up their development workflows, and creatives experimenting with new ways to generate content.
But beneath these visible transformations lies a subtler shift, one that’s taking place not in what we write or how we work, but in how we speak. And we might not even realize it’s happening.
The AI Echo in Our Everyday Speech
Hiromu Yakura, a postdoctoral researcher at the Max Planck Institute for Human Development in Berlin, first became aware of this phenomenon through personal experience. About a year after ChatGPT’s release, he noticed a curious change in his own vocabulary. Words like delve began appearing more frequently in his speech, a term he didn’t consciously use often before.
That small realization led to a much bigger question: Was this a personal quirk, or part of a larger trend?
Together with fellow researchers, Yakura launched a study to explore whether large language models (LLMs), such as the one powering ChatGPT, were not only altering written language, but also influencing the way people speak.
Tracing the Linguistic Footprint of AI
To investigate this idea, the team designed a two-phase approach. First, they prompted ChatGPT to revise millions of pieces of text, from emails and essays to academic papers and news articles, asking it to improve clarity or polish the writing. They then extracted a list of words that the model frequently added during this editing process. Words like realm, delve, and meticulous stood out. The researchers labeled these high-frequency additions as “GPT words.”
Next, they turned to a massive dataset of over 360,000 YouTube videos and 771,000 podcast episodes, analyzing spoken language from both before and after ChatGPT’s release. By comparing the usage of GPT words to “synthetic controls”, alternative synonyms not favored by the model, they were able to detect a clear upward trend in the use of GPT-influenced language.
A Cultural Feedback Loop in Action
This phenomenon reveals more than just a few quirky word choices. It suggests that a cultural feedback loop is beginning to form between humans and AI. We train AI models on our collective written language. These models then generate statistically remixed responses using patterns they’ve learned. As we use these tools and consume their outputs, we unconsciously begin to adopt some of these patterns ourselves.
Levin Brinkmann, a co-author of the study and fellow researcher at the Max Planck Institute, describes it this way: “The patterns that are stored in AI technology seem to be transmitting back to the human mind.”
The process is subtle, but its implications are significant. Language doesn’t just reflect how we think; it shapes how we think. If we begin to consistently echo the linguistic preferences of a machine trained on curated datasets, we risk losing the diversity and richness of human expression.
The Power of Authority and Perception
Another key factor in this shift is trust. We naturally imitate people, or systems we consider to be credible or intelligent. As AI tools become embedded in our daily routines and take on roles traditionally reserved for experts, their linguistic style begins to carry authority.
Brinkmann notes that people are more likely to mimic speech they perceive as competent or authoritative. As more individuals look to AI-generated content for answers, explanations, and insights, the style in which that content is delivered becomes a linguistic standard of sorts.
The result is not just individual word adoption but a broader change in tone, sentence structure, and even how ideas are presented in conversation.
Why It Matters
At first glance, this trend might seem harmless, even amusing. After all, what’s the problem with people using words like meticulous or realm more often? But the issue is not about the specific words. It’s about what they represent.
Language diversity is a vital part of cultural identity. It’s how communities maintain nuance, preserve history, and express individuality. If AI-generated language begins to crowd out natural linguistic variation, we could see a narrowing of communication styles, perspectives, and even worldviews.
This is why scholars like James Evans, a professor of sociology and data science at the University of Chicago, stress the importance of tracking AI’s cultural influence. Evans, who was not involved in the study, notes that analyzing word distribution is an effective way to understand early trends. But as LLMs become more advanced and their outputs more natural-sounding, these changes may become harder to detect.
To fully understand the long-term impact, researchers will need to go beyond vocabulary and study shifts in syntax, tone, and conceptual framing.
Looking Ahead
The influence of AI on culture is no longer a hypothetical; it’s already happening. Just two and a half years into ChatGPT’s existence, we’re seeing measurable changes in the way people speak. And this is likely just the beginning.
“Word frequency can shape our discourse or arguments about situations,” Yakura explains. “That carries the possibility of changing our culture.”
The intersection of AI and human communication is a space we need to observe closely, not just from a technical or ethical perspective, but from a cultural one. Because as AI continues to evolve, so too will the ways it subtly rewires our minds, one word at a time.
What are you noticing?
Personally, I’ve found “seamless” showing up almost everywhere, in emails, marketing copy, even casual conversations. It’s definitely one of those “AI words” that’s spreading fast.