On the Blandification of Writing

|

,

A recent NBC News article highlights new findings from Google and leading universities suggesting that AI may be making writing more uniform or blandification, such as changing in voice, tone, and meaning. Reporter Jared Perlo explains that researchers looked at how AI shapes responses to questions about happiness. They found that heavy use of large language models can influence not just writing style, but also the substance of people’s arguments.

The study reportedly examined three AI systems: Claude 3.5 Haiku, GPT-5 Mini, and Gemini 2.5 Flash. Researchers defined “heavy AI users” as those who generated more than 40 percent of their text using these tools.

Natasha Jaques, one of the lead authors of the study and a computer science professor at the University of Washington, told Perlo that heavy reliance on AI systems can lead to what she describes as the “blandification” of writing.

The LLMs are pushing the essays away from anything that a human would have ever written. They just change human writing in a way that’s very large and very unlike what humans would have done otherwise.

Key highlights of the study were:

  • Heavy use of large language models led to more neutral responses when discussing topics such as the link between money and happiness. It also made writing less personal and more formal, with fewer anecdotes and fewer real-life references.
  • Users felt that AI-assisted essays had less creativity and a weaker personal voice, even though their overall satisfaction with the essays remained similar.
  • LLMs often made larger and more noticeable edits than human editors, sometimes replacing significant portions of the original text.
  • The growing use of AI is raising important questions about how academic work should be evaluated, especially for acceptance at AI conferences.

This research highlights that LLMs are not able to adhere to peoples’ preferences and personalize how the human would have written the essay. An ideal LLM should write the essay that you would have written and just save you time. It’s not doing that at all. It’s writing a very different essay.

If you’re training a model on human feedback, the model has no boundary or perception of the difference between satisfying the humans and actually altering the human to make their preferences easier to satisfy.

Humans care about clarity, relevance, and impact, while AI cares about scalability and reproducibility. It’s changing our conclusions in ways that are already affecting our existing institutions.

– Natasha Jaques, a senior research scientist at Google DeepMind (Source: NBC News)

“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.

Thomas Juzek, a professor of computational linguistics at Florida State University, not involved in the study, told NBC:

  • This is a really good paper. What really struck me is this kind of illusion of using LLMs to perform a grammar check. This research shows that while a user might think they’re just doing a simple language check, the model is doing so much more.
  • Going forward, what does this mean for thought, language, communication, and creativity?

Prof. Jaques told NBC that she avoids using AI to write, instead using LLMs and their shortcomings as inspiration. “Sometimes, I’ll put a crappy version of what I’m trying to say in a conversational style into an LLM. That usually produces something which then motivates me to write it myself.”

Acknowledgement: This summary was generated using AI writing tools as part of my AI Explorations process. For more information, please refer to the original articles listed below.

Related Resource Links