Culture
Does AI kill creativity?
Creative tools can flatten output or expand iteration depending on use.
"AI kills creativity."
What this page actually tests
When generative AI substitutes for human ideation, it measurably homogenizes ideas, reduces collective diversity, and encourages dependence.
Wording note: Kills creativity implies a universal and permanent loss. The measurable concern is reduced diversity or weaker independent ideation under particular forms of use.
Creative homogenization is a real risk.
Confirmed. Experiments find that generative AI can improve some individual outputs while reducing collective diversity and independent idea variation. The effect depends on whether AI assists or replaces human ideation.
Why people repeat it
The concern is common because generated work often shares recognizable patterns, similar tools can steer many users toward similar ideas, and large volumes of inexpensive output can crowd creative markets.
What the sources support
Fact: Doshi and Hauser found access to generative AI ideas increased evaluated story creativity for some writers, especially less creative writers, while AI-assisted stories became more similar to each other.
Baseline: Individual output quality and aggregate creative diversity are different baselines. A tool can help one person and still make a category feel more samey.
Evidence conclusion: The evidence confirms that AI assistance can improve individual evaluations while homogenizing work across a group, making creative diversity a real tradeoff.
Source: Generative AI enhances individual creativity but reduces the collective diversity of novel content
Fact: The Copyright Office says AI used as an assistive tool does not automatically prevent copyright protection for human-authored parts of a work.
Baseline: Creative law already distinguishes human expression, tools, arrangement, editing, and uncopyrightable material.
Evidence conclusion: The evidence supports the idea that human creative control still matters when AI is part of the workflow.
Source: Copyright and Artificial Intelligence, Part 2: Copyrightability
Fact: The same copyright guidance treats purely AI-generated material without sufficient human authorship differently from human-authored expression.
Baseline: Accepting a first output is not the same as directing, selecting, revising, arranging, and transforming material.
Evidence conclusion: The conclusive caveat is that AI can reduce human creativity when it replaces decision-making instead of supporting it.
Source: Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence
Fact: A 2025 Nature Human Behaviour reanalysis of randomized brainstorming experiments found ChatGPT increased average idea creativity while decreasing the diversity of the combined idea pool.
Baseline: A stronger average idea and a broader set of ideas are different outcomes; brainstorming often needs both quality and variety.
Evidence conclusion: The repeated individual-gain and group-homogeneity pattern is a real caveat, but it still does not support saying creativity simply dies whenever AI is used.
Source: ChatGPT decreases idea diversity in brainstorming
Fact: A 2026 Nature Human Behaviour study compared 9,198 people with 215,542 language-model observations and found human divergent creativity was slightly higher on average, with greater variation and a stronger high-creativity tail.
Baseline: That large comparison tests humans and models on an established divergent-creativity task, not the full lived process of making art, revising work, or collaborating.
Evidence conclusion: The evidence rejects both easy extremes: current models do not erase human creative advantage, and their measurable creative performance is not zero.
Source: A large-scale comparison of divergent creativity in humans and large language models
Source balance
Checked both sides before calling it.
Supports the claim
- ChatGPT decreases idea diversity in brainstorming - AI-assisted brainstorming reduced diversity across the combined pool of ideas.
- Generative AI enhances individual creativity but reduces the collective diversity of novel content - The study finds reduced aggregate diversity in some AI-assisted creative outputs.
- Copyright and Artificial Intelligence, Part 2: Copyrightability - The Copyright Office emphasizes human authorship when evaluating AI-assisted creative work.
Challenges or narrows it
- A large-scale comparison of divergent creativity in humans and large language models - Humans retained a slight average advantage and a stronger high-creativity tail, while models still showed measurable divergent creativity.
- Generative AI enhances individual creativity but reduces the collective diversity of novel content - The same study finds individual creativity evaluations can improve with AI assistance.
- Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence - AI-assisted work can include human-authored elements when the human contribution is sufficient.
Baseline context
- A large-scale comparison of divergent creativity in humans and large language models - Provides a large same-task comparison of human and model divergent-creativity distributions.
- Copyright and Artificial Intelligence, Part 2: Copyrightability - Provides human authorship and assistive-use context.
- Generative AI enhances individual creativity but reduces the collective diversity of novel content - Provides individual-versus-aggregate creativity comparison.
Assessment: The core concern is confirmed: AI can homogenize output and reduce collective diversity under substitutive or convergent use. Individual performance gains define the tradeoff and conditions rather than disproving the effect.
Where critics may still have a point
- If everyone uses similar prompts and accepts similar outputs, creative diversity can shrink.
- AI spam can flood markets and make human work harder to find.
- Credit, compensation, and disclosure norms still need work even when AI is only part of the process.
Creative homogenization is a real risk.
The evidence points in both directions without canceling itself: some people produce stronger individual work with AI, while groups can converge on more similar ideas. Creative control, process design, attribution, and whether people continue generating their own alternatives determine the result.
Why this verdict: Peer-reviewed experiments directly support the bounded creativity claim that substitutive AI use can reduce collective diversity and idea variation even when it improves some individual evaluations.
How this was confirmed: Independent peer-reviewed experiments on fiction, brainstorming, and divergent creativity support homogenization and dependence risks. The cited experiments that found improved individual output provide the challenge baseline; the verdict confirms a conditional effect rather than claiming all AI use destroys creativity.
Sources
-
Generative AI enhances individual creativity but reduces the collective diversity of novel content
Used for: Evidence on individual creativity gains and reduced aggregate diversity.
-
Copyright and Artificial Intelligence, Part 2: Copyrightability
Used for: Human authorship, assistive use, and copyrightability distinctions.
-
Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence
Used for: Treatment of AI-generated and human-authored elements in creative work.
-
ChatGPT decreases idea diversity in brainstorming
Used for: Independent cross-check of the individual-creativity gain and collective-diversity tradeoff across brainstorming experiments.
-
A large-scale comparison of divergent creativity in humans and large language models
Used for: Large human-versus-model comparison showing a slight average human advantage and a stronger high-creativity tail.