People say “delve” far more often than they used to, and the timing points straight at ChatGPT, the AI chatbot. Researchers tracked the words ChatGPT overuses, then counted those words across hundreds of thousands of podcasts and recorded talks. Their study is among the first to track spoken language this way, and the jump it found came right after the tool went live.
Adam Aleksic, who studies where words come from, sees that one word as a warning sign for something bigger. In a TED talk, he argues that the tools people type into and scroll through are not neutral. They are changing how people talk, and how people talk then feeds back into how they think.
By Aleksic’s best guess, the chatbot’s fondness for “delve” goes back to how it was trained. Part of that training was handled by workers in Nigeria, where the word is common in daily speech. Their habit slipped into the system, and because a model copies whatever it sees most, ChatGPT leaned on the word even harder than those workers would. A small quirk, repeated across enough text, becomes the machine’s default way of talking.
Then a loop sets in. People read the chatbot’s wording, pick up a little of it and type it into their emails and posts. That writing becomes training material for the next version of the model, which serves the same wording back to its users. With every loop, everyday speech drifts closer to how the software writes.
Aleksic’s defense is one question, asked often, about why a person is seeing something, saying it or thinking it, and why the platform wants them to. He pairs the question with simple habits, such as reading beyond the usual sources and talking to people who sound nothing like the feed. He also suggests catching word choices before they turn automatic.
Thinking is where the effect runs deepest, because borrowing a machine’s words means slowly borrowing its sense of what is worth saying. Letting TikTok or Spotify pick what to watch or hear works the same way. As more people pick up the same phrasing, the quirks that make voices distinct flatten out, so slowly that the change feels like personal taste.
