AI Chatbots Tend Towards Conformity, Potentially Harming Human Creativity

AI chatbots are often promoted as tools to boost creativity, not to subtly diminish it. However, new studies indicate that this narrowing effect might be precisely what happens when we rely too heavily on them.

Research published in Engineering Applications of Artificial Intelligence reveals that prominent AI models like Gemini, GPT, and Llama frequently converge on similar conceptual frameworks when given creative assignments. While individual responses might seem innovative and useful, a different pattern emerges upon broader examination: outputs start to overlap across numerous prompts and users.

Investigators compared human participants against a variety of AI models using standard creativity assessments, such as brainstorming novel uses for everyday items or generating lists of unrelated words. The AI performed well individually, but as a collective, its ideas demonstrated significantly less diversity.

Similar Bots, Identical Patterns

The research team evaluated over 20 different models from various companies against more than 100 human participants. The findings remained consistent: AI-generated responses exhibited a much narrower range, even when the models originated from distinct foundational architectures.

When analyzed for similarities, chatbot outputs clustered closely together, whereas human responses explored a considerably broader intellectual landscape.

This consistent pattern was observed across different tasks. Whether the goal was to produce ideas or abstract concepts, the models consistently drew upon familiar structures and repetitive phrasing.

Efforts to encourage greater variety yielded limited success. Increasing the degree of randomness offered some improvement but quickly compromised coherence. Prompting the AI for more imaginative results provided only a marginal shift, failing to significantly broaden the scope of ideas.

The Impact on Your Originality

Superficially, AI can still appear impressive, with many of its responses matching or even surpassing the average human for originality.

The problem becomes more apparent at scale. When many individuals use the same AI tools for brainstorming or writing, they frequently draw from the same underlying computational patterns. Over time, this inherently restricts the range of available ideas, even if each individual output seems distinct.

A key limitation stems from what these systems lack: personal context, lived experiences, or genuine intent. This deficiency might inherently limit the divergence of their ideas, regardless of the prompt's sophistication.

There's also a behavioral aspect to consider. The study suggests that people might excessively depend on AI suggestions rather than expanding on their own thoughts. This reliance could further reduce the diversity of human ideas over time.

Future Considerations

This issue doesn't seem to be confined to a single product; it appears to be a shared characteristic across contemporary AI systems. Even models from different developers produced overlapping results, suggesting a fundamental constraint in how these tools generate creative content.

For now, AI functions best as a starting point, not a definitive solution. Utilize it to spark initial directions, then develop and expand upon those ideas independently. Otherwise, you risk merely reassembling the same concepts as everyone else, rather than genuinely innovating.