1. The Homogenization Problem
A 2025 study examining GPT-4 and collective creative diversity found that human-written essays contributed substantially more new semantic diversity than additional GPT-4-generated essays. The finding does not mean that humans were individually several times more creative than AI. It concerned collective diversity: how much new variation each additional contribution added to the overall pool.
This creates a paradox. AI can make individual outputs more polished, coherent and effective while potentially making a collection of outputs more similar.
2. But People Use AI Differently
The important question is therefore not simply whether AI increases or decreases creativity. It is how people use the same AI.
A predictive user asks AI for the answer and accepts the most useful response. This is efficient, but it naturally pushes thinking toward the answer the machine considers most probable.
A productive user starts with an idea and asks AI to improve, challenge, test or extend it. The human supplies the direction while AI increases the speed and scale of iteration.
A generative user goes further. They explore unusual connections, move between different fields, test strange possibilities and follow ideas before knowing where they will lead.
The same machine is therefore being used in three different ways.
3. Prediction, Production and Generation
The distinction can be summarized simply: predictive use tends toward convergence, productive use increases iteration, and generative use can increase exploration and divergence.
The technology has not changed. The cognitive strategy of the user has.
This may be one of the most important variables in understanding AI’s effect on human thinking.
4. AI as a Sea of Possibilities
AI can be imagined as a huge sea of patterns, knowledge and associations.
One person may take a cup from it. Another may cast a net. Someone else may dive deeply and search for something that nobody was specifically looking for.
The resource is the same, but the extraction strategy is different.
That is why neither “AI will destroy creativity” nor “AI will make everyone more creative” is sufficient. What people do with the technology may matter enormously.
5. The Possible Future Divide
The future divide may therefore not simply be humans versus AI.
It could become a difference between people who use AI predictively and people who use it generatively.
Predictive users may increasingly converge toward machine-mediated answers. Generative users may use the same machines to investigate more possibilities, connect more domains and develop ideas at a much larger scale.
The second proposition is still a hypothesis requiring further testing. Existing research demonstrates concerns about collective creative homogenization, but it does not yet establish that highly generative individuals systematically benefit more from AI.
The deeper question is therefore not simply: “Does AI make people think alike?” It may be: “What happens when different kinds of thinkers use the same intelligence differently?”
Read the full analysis:
Read the full analysis: The Same AI. Different Minds.
Source: Moon, K., Green, A. E., & Kushlev, K. (2025), “Homogenizing effect of large language models (LLMs) on creative diversity: An empirical comparison of human and ChatGPT writing.”