AI is not killing creativity in the simple, cinematic sense. It is not arriving like a villain, switching off imagination, and leaving culture in the dark. The more unsettling reality is less dramatic and more plausible. AI is making creation faster, cheaper, smoother, and more available, while also increasing the risk that large amounts of creative work start to sound, look, and feel eerily similar. That is not the death of creativity. It is the industrialization of it. And because generative AI is already mainstream, the stakes are no longer theoretical. The OECD says more than one-third of individuals across OECD countries used generative AI tools in 2025.
The wrong fear and the right one
The wrong fear is that AI will eliminate human creativity altogether. Human creativity is too bound up with memory, taste, biography, obsession, embarrassment, desire, contradiction, and risk for that. The right fear is that AI will reward the most frictionless version of making things: the version with fewer dead ends, fewer strange detours, fewer original failures, and fewer deeply personal signatures. Creativity does not disappear first by being banned. It disappears by being optimized into predictability. That is why the current debate is often misframed. The question is not whether AI can generate content. It plainly can. The question is what happens to originality, cultural diversity, and artistic courage when millions of people begin making work through the same statistical machinery.
Why AI feels creative even when it often compresses difference
Generative AI feels creative because it is excellent at recombination. It can absorb huge patterns of style, structure, rhythm, genre, and convention, then produce outputs that look polished and plausible at extraordinary speed. For many tasks, that already beats the blank page. In a 2024 Science Advances study, access to generative AI ideas made short stories more creative, better written, and more enjoyable on average, especially for less creative writers. But the same study found a collective downside: AI-assisted stories became more similar to each other, including a measured increase in similarity among AI-aided outputs.
That dual result matters. AI often improves the local experience of making something. It gives momentum, structure, confidence, and useful scaffolding. Yet the very mechanism that makes it so helpful also explains why it can flatten the landscape. These systems are trained to produce likely, legible, coherent continuations. They are very good at moving toward what fits. Originality, however, often begins where “what fits” breaks down. It begins with odd judgments, non-obvious synthesis, tension, refusal, or a voice that does not immediately optimize for readability and consensus. The danger is not that AI makes everything bad. The danger is that it makes an enormous amount of work competent in the same direction.
The creativity boom for individuals
This is the part critics often miss. AI really does expand access to creative work. It lowers the threshold for people who struggle with first drafts, composition, ideation, formatting, structure, or technical execution. For many users, that is not a trivial convenience. It is the difference between starting and not starting. In the story-writing experiment, less creative writers benefited the most from AI assistance, gaining substantial improvements in judged writing quality and enjoyability. The tool narrowed the gap between weaker and stronger creators. That is a genuine democratizing effect.
Research on open-ended problem-solving points to a similar conclusion. Harvard Business School’s coverage of an Organization Science paper reported that humans contributed more novel suggestions, AI produced more practical solutions, and some of the strongest outcomes emerged when people and machines worked together. That is an important clue. AI does not need to be a substitute for creativity to reshape creative work. It only needs to become a powerful collaborator, one that handles certain forms of exploration, drafting, or constraint-solving faster than humans can on their own.
There is another reason the “AI kills creativity” claim is too blunt. In some domains, generative tools expand the creative frontier by increasing the number of attempts people can make. A 2025 Science Advances paper on text-to-image AI found that these tools can increase novel contributions through high-volume exploration, first concentrating breakthroughs among some users and later broadening participation across a wider group. In other words, AI can help more people search a creative space more rapidly. That is real creative leverage. But it is not the same thing as saying the machine itself has solved originality. The gain often comes from scale and speed, not from replacing human taste or meaning.
The monoculture risk
The strongest case against complacency is not that AI outputs are always mediocre. It is that they are often homogenizing at scale. A 2025 study on “creative homogeneity across LLMs” found that responses from multiple large language models were far more similar to each other than human responses were to one another, even after controlling for response structure. The models could appear individually creative while remaining collectively narrow. That is precisely the kind of pattern a culture industry should worry about.
Once that pattern meets mass adoption, the effects compound. If millions of writers, designers, marketers, students, agencies, and brands start with similar prompts, similar models, similar defaults, similar optimization habits, and similar pressure to publish quickly, then the output stream does not just expand. It converges. The oddity gets sanded down. The middle thickens. The surface quality rises while the variance shrinks. UNESCO has explicitly identified risks around copyright, skills and jobs, fairness and transparency, cultural diversity, and the long-term market impact of generative AI on cultural and creative sectors. That concern is not aesthetic snobbery. It is structural. Diversity in culture depends on more than volume. It depends on difference.
This is why the phrase “AI-generated creativity” can be misleading. It treats the main issue as whether the output looks inventive. But cultural health is not measured only by whether one image, one song, or one paragraph seems clever in isolation. It is measured by whether a creative ecosystem still produces disagreement, distinct sensibilities, local styles, minority voices, difficult forms, and new aesthetics that are not simply recombinations of already dominant patterns. A culture can be flooded with content and still become less creatively alive.
Taste becomes the scarce skill
As generative tools improve, the scarce skill shifts. It becomes less valuable to produce a clean generic output on command, because machines can increasingly do that. What becomes more valuable is deciding what should exist, what should be rejected, what deserves further development, what feels false, what feels derivative, and what still carries a human stake. The creative act moves upstream and downstream: framing, selecting, editing, combining, and refusing. Taste is no longer the decorative layer after production. It becomes the production advantage itself. This fits the hybrid pattern seen in research, where humans and AI often perform best not when the machine is treated as an oracle, but when it is treated as a fast, unruly assistant whose outputs still require judgment.
That has legal echoes too. The U.S. Copyright Office’s recent AI work has kept human authorship and creative control at the center of copyrightability analysis for AI-assisted outputs. That does not settle every legal fight around AI and culture, but it reflects a broader principle that extends beyond law: society still distinguishes between automated generation and human-directed authorship. We instinctively understand that prompting is not identical to composing, and that selecting among machine outputs is not always the same as originating a work’s expressive core.
What AI still cannot do for you
AI can give you options. It cannot give you a life. It can imitate a voice. It cannot have a past. It can simulate conviction. It cannot risk reputation, love, humiliation, or belief. These are not mystical objections. They are practical ones. The works people return to over years are rarely memorable because they are merely fluent. They matter because they carry pressure from somewhere real: a scene lived through, a conflict endured, a private obsession sharpened into form, a moral risk taken in public, a sensibility no committee would have approved in advance.
That is why the best human work may become more visible, not less, in an AI-saturated environment. Once competent output becomes abundant, the premium shifts toward work that bears unmistakable marks of selection, experience, and intent. Readers, viewers, and listeners may not always articulate it in those terms, but they can often feel the difference between something assembled from the center of the distribution and something made by a mind that is not trying to sound like everyone else. In a world full of generated fluency, the human signature gets easier to notice.
How creative people should use AI without letting it flatten them
The practical answer is not abstinence. It is discipline. Use AI to expand the search space, not to collapse it too early. Use it to produce alternatives, tensions, counterexamples, structures, and research leads. Use it to get past friction, not to outsource judgment. The moment a creator lets the model decide what is “good enough,” the model’s center of gravity starts replacing the creator’s own. Research on text-to-image systems suggests that AI can widen exploration through sheer output volume. That can be useful. But volume only helps if someone is still curating with standards that are not generated by the same system.
The strongest creators will probably be the ones who make AI work against itself. They will use it to generate the obvious version and then move away from it. They will ask for opposites, contradictions, uglier drafts, stranger references, narrower constraints, more local detail, more lived texture. They will protect parts of the process that should remain slow. They will preserve zones where uncertainty, boredom, patience, and revision still matter, because those are often the places where voice forms. AI should handle repetition, formatting, synthesis, and permutation. It should not be allowed to silently standardize imagination.
Creativity survives by getting harder again
So is AI killing creativity? Not exactly. It is doing something more ambiguous and, in some ways, more dangerous. It is making creativity easier to perform and harder to distinguish. It raises the baseline, lowers the barrier, expands participation, and accelerates experimentation. It also threatens to reward sameness, dilute cultural variance, and reduce the amount of genuinely surprising work that reaches the surface.
The future of creativity will not be decided by whether AI can generate impressive outputs. That question is already old. The real test is whether human creators, editors, audiences, schools, studios, and platforms still know how to value work that resists the default. If they do, AI may become a powerful tool inside a richer culture. If they do not, culture may become smoother, faster, and more crowded while growing less original underneath. That is the sharper diagnosis. AI is not killing creativity. It is forcing creativity to prove that it is more than efficient production.
Sources
Generative AI enhances individual creativity but reduces the collective diversity of novel content
Peer-reviewed Science Advances study on how AI assistance can raise judged creativity for individuals while reducing diversity across outputs.
https://www.science.org/doi/10.1126/sciadv.adn5290
AI found to boost individual creativity – at the expense of less varied content
University of Exeter summary of the 2024 Science Advances experiment, including accessible reporting on gains for less creative writers and the rise in similarity across AI-aided stories.
https://www.eurekalert.org/news-releases/1050888
We’re Different, We’re the Same: Creative Homogeneity Across LLMs
2025 study examining standardized creativity tests across multiple LLMs and finding much lower population-level variability than in human responses.
https://arxiv.org/abs/2501.19361
Can AI Match Human Ingenuity in Creative Problem-Solving?
Harvard Business School article summarizing research showing humans contribute more novel ideas, AI produces more practical ones, and hybrid collaboration can be especially strong.
https://www.library.hbs.edu/working-knowledge/generative-ai-and-creative-problem-solving
Who expands the human creative frontier with generative AI
Accessible full-text version of the 2025 Science Advances paper on text-to-image AI, productivity effects, and how AI-assisted exploration can broaden novel contributions over time.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12407059/
Art and the science of generative AI
Science commentary on how understanding changes in creative work should guide AI’s impact on the media ecosystem.
https://www.science.org/doi/10.1126/science.adh4451
Copyright and Artificial Intelligence
U.S. Copyright Office overview of its multipart AI report, including Part 2 on copyrightability of outputs created using generative AI.
https://www.copyright.gov/ai/
Works Containing Material Generated by Artificial Intelligence
U.S. Copyright Office guidance for registration of works containing AI-generated material, relevant to human authorship and disclosure.
https://www.copyright.gov/ai/ai_policy_guidance.pdf
The impact of AI and the digital transformation on the CCSI
UNESCO policy monitoring page outlining key risks for cultural and creative sectors, including cultural diversity, fairness, jobs, and copyright-related concerns.
https://www.unesco.org/creativity/en/policy-monitoring-platform/impact-ai-and-digital-transformation-ccsi
Generative AI
OECD overview page with current adoption data and policy context for generative AI use across OECD countries.
https://www.oecd.org/en/topics/generative-ai.html
Author:
Jan Bielik
CEO & Founder of Webiano Digital & Marketing Agency
