At 5:48 PM (KST) on August 18, 2026, a post by the X account @HowToPrompt__ passed 177,000 views. The key sentence was: "Students without LLM access were 8x more creative than students with access." It said the researchers called this the "homogenization effect."
8x Student Creativity: Was It Real?
The original post said it "compared students with and without LLM access." It added that changing prompts or parameters to make the AI more creative didn't help. The numbers and wording are specific, which makes the sentence tempting to share as is.
The Error a Community Note Flagged
An X Community Note was attached to the same post. It corrected: "This study didn't compare students with and without LLM access; it compared essays written by humans and essays written by AI." The actual paper backs this up. "Homogenizing effect of large language models (LLMs) on creative diversity", by Kibum Moon, Adam E. Green, and Kostadin Kushlev of Georgetown University, was published in Computers in Human Behavior: Artificial Humans, vol. 6 (2025), article 100207. It compared 2,200 college application essays split into those written by humans and those written by GPT-4. Whether students had access wasn't a variable in this study to begin with.
What the Diversity Growth Rate Says
The metric the paper actually measured is the "diversity growth rate." It looks at how many new ideas are added each time you add one more essay. According to the original abstract of the paper, each additional human-written essay added more new ideas than each additional GPT-4 essay, and the gap grew as the number of essays increased. Even when prompts and parameters were adjusted to make the AI's output more varied, the gap didn't go away. The "8x" multiple appears nowhere in this abstract.
What Happens When a Team Hands Every Draft to AI
This study doesn't say how good any one person's piece of writing is. It describes a group-level phenomenon: when many people each pull a first draft with AI, the individual outputs may be polished, but the team's output as a whole starts to look alike. It's worth thinking through especially when several people work in parallel with the same tool, such as brainstorming, copy drafts, and planning documents. Even if you use AI drafts as a starting point, a workflow that deliberately keeps a step where each team member writes from their own perspective is closer to the practical implication this study points to.
Checked Through the Abstract, Unverified Beyond That
This article compared the original at the level of the paper's abstract and introduction. It didn't check the statistical methodology across the full text of the paper. I also looked at a point from @risaiks in the replies to the same post, but treated it only as a user opinion, not a verified primary source. The "8x" wording also isn't new. Around 2025, researcher Valerio Capraro also cited the same 2,200-essay dataset and used the phrase "students without LLM access were 2 to 8 times more creative", which suggests the same study has been paraphrased in different ways several times. You can see the Community Note and all the replies in the original @HowToPrompt__ post thread.