AI Chatbots Converge on Similar Ideas, Limiting Creative Diversity
Key Points
- Research published in Engineering Applications of Artificial Intelligence examined AI creativity.
- More than twenty chatbot models, including Gemini, GPT and Llama, were tested.
- Over one hundred human participants provided a baseline for comparison.
- AI outputs clustered tightly, while human ideas covered a broader space.
- Increasing randomness added variety but reduced coherence.
- Prompting for imagination nudged results but did not meaningfully expand diversity.
- Widespread reliance on similar AI tools may compress overall idea diversity.
- Chatbots lack lived experience, intent, and personal context.
- Authors recommend using AI as a starting point, not a finish line.
A study published in Engineering Applications of Artificial Intelligence finds that leading AI chatbots such as Gemini, GPT and Llama often generate overlapping ideas when tasked with creative problems. Testing more than twenty models from various companies against over one hundred human participants, researchers observed that AI outputs clustered tightly while human responses covered a much broader space. Efforts to increase randomness or prompt the models for greater imagination produced only modest gains and often reduced coherence. The findings suggest that while AI can produce impressive individual suggestions, widespread reliance on these tools may compress the overall diversity of ideas.
Study Overview
A recent research paper in the journal Engineering Applications of Artificial Intelligence examined how modern AI chatbots perform on standard creativity tests. The study compared the output of more than twenty models from different companies—including well‑known systems such as Gemini, GPT and Llama—to the responses of over one hundred human participants.
Key Findings
Individually, many AI responses appeared original and useful, often matching or even surpassing the average human answer in originality. However, when the results were aggregated, a clear pattern emerged: the AI-generated ideas occupied a much tighter conceptual space. Visual mapping of similarity showed chatbot answers clustering closely together, whereas human responses spread across a far wider area.
The convergence persisted across different creative tasks, from brainstorming new uses for everyday objects to listing unrelated words. Even models built by different organizations produced overlapping outputs, indicating a shared limitation in how these systems generate ideas.
Attempts to Broaden AI Creativity
Researchers experimented with increasing randomness in the models and prompting them to be more imaginative. While higher randomness introduced a slight increase in variety, it quickly compromised coherence. Prompting the AI for greater imagination nudged results in the right direction but did not meaningfully widen the range of ideas.
Implications for Users
The study highlights a potential risk when many people rely on the same AI tools for brainstorming or writing. Because the models draw from similar underlying patterns, large‑scale usage can compress the overall diversity of ideas, even if each individual suggestion seems distinct.
AI systems lack lived experience, intent, and personal context, factors that may limit how far their ideas can diverge. Additionally, the research suggests that users may lean too heavily on AI suggestions instead of extending their own thinking, further reducing idea diversity over time.
Practical Takeaway
According to the authors, AI chatbots work best as a starting point rather than a final solution. Users are encouraged to treat AI‑generated suggestions as sparks for further development, ensuring that human creativity remains the driving force behind original ideas.