AI Homogenization: How Large Language Models Are Flattening Linguistic Diversity and Personal Expression

Large Language Models Are Reducing Linguistic Diversity Across Digital Content

Research published in Nature Human Behavior demonstrates that AI homogenization is fundamentally altering how humans write and express themselves. Zhivar Sourati, a PhD student at the University of Southern California, analyzed approximately 80,000 academic papers, 400,000 news articles, and 300,000 Reddit stories written before and after ChatGPT's November 2022 launch1

1

2

2

. The findings reveal a troubling trend: after the adoption of AI writing tools, variation in writing complexity increasingly converged toward common stylistic norms across all three datasets2

2

.

More than a third of all internet web pages published since ChatGPT's release were authored by AI, according to the Pew Research Center2

2

. This flood of AI-generated language isn't just changing what we read—it's reshaping how humans write. Sourati's team conducted experiments where GPT-3.5, Gemini, and Llama 3 rewrote human-authored texts using various prompts. While the models preserved meaning, they reduced variation in writing complexity by 21% to 50%2

2

.

AI Killing Personal Expression and Cultural Nuances

The phenomenon resembles what researchers call "McDonaldization"—invoking how characteristics of the fast-food industry, such as efficiency and predictability, have influenced society and led to more uniformity1

1

. But large language models differ from previous technologies that merely spread information. They actively shape it, flattening language use, creativity, and cultural values1

1

.

Emily Wenger, a computer scientist at Duke University, warns that cultural and cognitive homogenization poses an existential risk. "If we're all using AI to write our e-mails or whatever, what do we become as a species?" she asks1

1

. Her research tested 22 LLMs and 102 people on three creative tasks, including brainstorming tasks that asked participants to name alternative uses for common objects. The LLMs produced slightly more original ideas than humans, but their responses were far more similar to each other than human responses were1

1

.

AI Making Us Think the Same Threatens Creative Diversity

Other research mapped narrative features of short stories written by people and five LLMs, analyzing plots, characters, settings, and other elements. AI-generated stories clustered together based on their narrative features, while human-written ones were more spread out1

1

. Working with AI tools directly reduces human creativity. In a 2024 study, people wrote eight-sentence short stories on assigned topics, with some first asking an LLM for ideas. AI use led to stories that evaluators rated as more novel and enjoyable, except for those by the most talented writers. But it also made stories more similar to each other1

1

.

A 2025 study involving brainstorming tasks found that ideas people submitted were more creative when they used ChatGPT, but AI use reduced the diversity of ideas1

1

. A meta-analysis published in April found that AI homogenization was strongest for idea generation, especially in complex or constrained tasks1

1

.

AI's Societal Impact on Identity and Cultural Expression

The implications extend beyond creativity to identity itself. Sourati's research on local news articles, arXiv preprints, and Reddit posts found that after ChatGPT's initial release, linguistic measures of writing style showed a decreased variation1

1

. When researchers used LLMs to correct grammar in human-written texts, they erased many signifiers of personality, moral values, and demographics1

1

.

Generative AI can mute cultural nuances between different populations. In one study, people in India and the United States described their favorite rituals, heroes, and symbols. Half wrote using an autocomplete tool that suggested up to nine words each time they paused. This AI use increased writing similarity between Indians, between Americans, and between the two groups, leading Indians to sound more American in their wording1

1

. Descriptions of the Indian festival of Diwali became more generic, losing rich cultural details1

1

.

The Risk of Groupthink and Long-Term Adaptability

Wenger warns we might fall prey to groupthink. "We have suffered as a society when we silence or ignore edge voices," she notes1

1

. Sourati imagines a worst-case outcome where humans collectively become less adaptable. "That's actually really scary for me," he says1

1

. The question researchers are now exploring: what can be done to combat AI homogenization before it fundamentally alters how we think, create, and express our individual and cultural identities?