AI writing assistants linked to decline in linguistic diversity

A study published in ‘Nature Human Behaviour’ by a team from the University of Southern California suggests that the widespread use of Large Language Models (LLMs) is causing a systemic decline in human linguistic diversity. By analyzing over 880,000 texts across platforms including Reddit, Patch, and arXiv, researchers found that the variance in writing complexity dropped significantly following the public release of ChatGPT in late 2022.

The research indicates that while AI tools are highly effective at maintaining semantic meaning—with 87% of rewritten texts retaining over 95% of the original meaning—they tend to flatten individual style. Because LLMs function by predicting the most statistically probable next word, they often replace unique or rare expressions with more conventional, predictable phrasing. This process can reduce text complexity variance by as much as 50%.

In response to these evolving dynamics, academic programs are beginning to address the implications of AI on cognition. A course offered by Circle U invites graduate students to critically examine how LLM assistance impacts epistemic agency, critical thinking, and the future of academic writing and reasoning.