AI Tools Reduce Human Writing Diversity by Up to 50% Study Finds
A comprehensive analysis published in Nature reveals that widespread AI adoption is driving human language toward predictable, standardized norms, reducing writing complexity by 21% to 50%.
As generative artificial intelligence becomes deeply integrated into daily writing, human language is growing increasingly predictable and uniform. A series of studies and analyses, including a comprehensive review published in the journal Nature, points to a clear trend: AI-based writing tools are nudging human expression toward safe, standardized, and average formulations.
According to the review, texts generated or assisted by large language models (LLMs) exhibit significantly higher similarity than those written solely by humans. AI rewriting tools consistently favor conventional phrasing over rare, creative, or highly specific expressions.
Analyzing Hundreds of Thousands of Texts
Researchers conducted a time-series analysis of a massive dataset spanning the period before and after the public launch of ChatGPT in November 2022. The corpus included:
-
Approximately 80,000 academic papers;
-
400,000 news articles;
-
300,000 stories and posts from the social media platform Reddit.
Across all three domains, the adoption of AI tools was followed by a sharp convergence in writing complexity toward shared norms. The variance in writing complexity dropped by 21% to 50%, indicating that diverse writing styles are rapidly being replaced by a centralized linguistic standard.
Furthermore, researchers noted a significant spike in the use of AI-associated vocabulary in academic abstracts within computational linguistics and neuroscience. In contrast, the impact on mathematics papers remained negligible.
The Rise of 'Robotoid Humanness'
This shift has led researchers to coin the term "robotoid humanness." The concept suggests that repeated exposure to AI systems conditions humans to gradually adapt their thoughts and self-expression into formats that are easier for machines to process. The ultimate result is a highly predictable and less diverse language.
According to Fast Company, a separate study focusing on scientific publishing found that while the number of papers published per author increased after the launch of ChatGPT, the linguistic style and content of these papers became significantly more homogeneous.
Beyond Style: The Threat to Creativity
Scientists warn that the implications of this linguistic convergence extend far beyond aesthetics.
"The convergence toward a uniform language could impact creativity, cultural expression, and the very way people formulate thoughts and adopt ideas. When identical phrasing and patterns are repeated, they blur personal identity and encourage groupthink."
Data presented in the review indicates that since November 2022, more than a third of all new web pages have been generated with the help of AI. This has simultaneously accelerated the development of AI detection tools, such as Pangram.
One of the key characteristics identified in AI-assisted writing is the tendency to repeatedly hammer home the same core idea, attempting to turn every single sentence into a highly structured, "actionable" conclusion. What initially appears to be an advantage of clear and organized writing ultimately results in a rigid uniformity that makes it difficult to distinguish one author from another.
