AI is making us sound the same—and killing our personal expression

More than a third of all internet web pages published since ChatGPT’s November 2022 launch were authored by AI, according to the Pew Research Center. And the writing is becoming simpler than ever. That flood of AI-generated language may also be changing the way humans write. A new paper in Nature Human Behavior finds that LLMs can push people’s writing toward a narrower set of linguistic norms, reducing stylistic diversity and obscuring some of the identity signals embedded in our words. Three years ago, lead researcher Zhivar Sourati read The Secret Life of Pronouns: What Our Words Say About Us, James Pennebaker’s book that explored how people’s personal values, backgrounds, and beliefs can lead them to communicate the same ideas using different words. At the same time, LLMs were becoming more popular, and Sourati began to wonder whether the tech might reduce that linguistic diversity. “We were kind of asking, what would actually happen when everyone’s text is going through the same models?” Sourati says. “If you have this layer on top of everyone’s text, would you change any of these artifacts that we know about people’s texts? Would it affect it in any significant way? What are…