Ever found yourself using words like “delve” or “meticulous” more often than you used to? You might not be alone. A new study suggests that in the wake of AI chatbots like ChatGPT, it’s not just the AI learning to sound more human—it seems even humans are starting to talk more like AI.
For a while, the debate has been whether we can spot AI-generated content by its particular style. But now, researchers from the Max Planck Institute for Human Development in Germany have found that our spoken language is subtly shifting. They observed a noticeable increase in what they’ve dubbed “GPT words” in human conversations. The trend started since ChatGPT hit the scene a couple of years ago.
Researchers find that people are talking more like ChatGPT
So, how did they figure this out? The researchers fed millions of pieces of human writing—emails, essays, and news stories—into ChatGPT. Then, they asked the AI-powered chatbot to “polish” the text. This helped them identify words that the AI seemed to favor. This list included words like “underscore,” “comprehend,” “bolster,” “boast,” “swift,” “inquiry,” and “groundbreaking,” in addition to “delve” and “meticulous.” The researchers then tracked the frequency of these words in over a million YouTube videos and podcast episodes from before and after ChatGPT’s launch.
The results were quite eye-opening: these “GPT words” have indeed become more prominent in spoken English. It’s almost as if a new “cultural feedback loop” has emerged. As one of the study’s co-authors, Levin Brinkmann, put it, “The patterns that are stored in AI technology seem to be transmitting back to the human mind.” He noted that we tend to imitate those we perceive as knowledgeable or important. It seems many people are now unconsciously viewing AI in that light.
AI influencing the evolution of language
Of course, language is always evolving, and many factors contribute to how we speak. This study primarily focused on specific versions of ChatGPT and included a good amount of academic content in its analysis. However, the findings offer a glimpse into how our interactions with advanced AI might even be shaping our own linguistic landscape. It’s a reminder that as AI becomes more integrated into our daily lives, its influence extends beyond just tools and into the very fabric of how we communicate.
This isn’t the first study on the effects of AI on human behavior. A report from a few weeks ago concluded that over-reliance on AI could be making us “dumber.” There are also studies warning about AI chatbots impersonating psychologists on platforms like Instagram.
