Study Finds Heavy AI Use Alters Meaning and Voice in Human Writing

Large language models are fundamentally changing not just the style but the substance of human writing, according to new research from the University of Washington that quantifies how AI assistance reshapes arguments and erases individual voice.

The study found that users who heavily relied on LLMs produced responses that diverged significantly in meaning from participants who used AI sparingly or not at all. When asked to revise human essays, the three leading AI systems made substantially larger edits than human editors in identical situations, altering the underlying meaning of the work.

"The LLMs are pushing the essays away from anything that a human would have ever written," said Natasha Jaques, a lead author and computer science professor at the University of Washington. "They just change human writing in a way that's very large and very unlike what humans would have done otherwise."

The research team documented a dramatic shift toward impersonal language: essays from heavy AI users contained 50 percent fewer pronouns and included fewer anecdotes and references to human experience. While human editors typically substituted individual words and preserved most original vocabulary, LLMs "replace a much larger fraction of the original writing than humans do when revising their own work," the paper states.

"This substitution of words contributes to the loss of individual voice, style, and meaning, as the unique lexical fingerprint of each writer is overwritten by the given model's preferred vocabulary," the authors wrote. The phenomenon, which Jaques termed "blandification," extends beyond cosmetic changes to alter the core arguments writers present.

(The study analyzed essays from a 2021 database predating widespread LLM adoption, examining differences in editing patterns and assessing how AI use affects peer review criteria at leading AI conferences.)

The findings arrive as industry observers debate the creative ceiling of language models. In a November 2025 interview, OpenAI CEO Sam Altman predicted LLMs would enable breakthroughs in climate science and physics, yet acknowledged that even future iterations "may only be able to produce something equivalent to 'a moderately good poem written by a real poet.'" Developers and researchers told The Atlantic that AI's lack of lived experience creates metaphors and descriptions that feel unnatural and lack emotional weight.

Meanwhile, the competitive landscape around AI writing tools continues to shift. Chinese open-source LLMs are gaining ground in app downloads, with estimates suggesting 80 percent of US tech startups now use Chinese open-source models, according to a US congressional advisory body. The open-source nature of code allows it to move instantaneously across borders, circumventing semiconductor export restrictions that were designed to maintain American technological leadership.