AI vs. Human Creativity—Who Wins?

The surprising answer isn’t one or the other. It’s that AI has reached average human creativity—but the most original minds still have a decisive edge.

Photo by Neeqolah Creative Works on Unsplash

I used to think creativity was uniquely human. Then I watched an AI generate a poem that made me feel something. Then I read another AI’s attempt at the same prompt, and another, and another, and slowly realized they were all starting to sound the same.

That’s when I stopped asking “Can AI be creative?” and started asking a more interesting question: What kind of creativity are we talking about?

After digging into the latest 2026 research—including the largest studies ever conducted comparing human and AI creative performance—I found an answer that surprised me. It’s not a simple win for either side. It’s a story about averages versus extremes, about what AI copies and what only humans can originate.

The Numbers That Changed My Mind

Let me start with the finding that stopped me cold. A massive study published in Scientific Reports tested over 100,000 humans against leading AI models on standardized creativity tasks. The result? GPT-4 now exceeds the average creative performance observed in humans on tests of divergent thinking—generating many different ideas from a single starting point—10.

On a scale where average human creativity scores around 79 out of 100, top AI models like GPT-4 score roughly the same—sometimes slightly higher. For the person on the street, AI has caught up.

But here’s where the story gets interesting. The same study found that the most creative humans still outperform even the best AI systems by a significant margin—10. The top 10% of human creators open a gap that AI hasn’t closed. When the researchers looked at creative writing tasks—haiku composition, movie plot summaries, and short stories—skilled human creators retained a clear advantage—10.

So AI beats the average human but loses to the exceptional one. That’s the first layer of nuance.

The Diversity Problem: AI Is Homogenizing Creativity

Here’s the finding that genuinely worries me. A Duke University study published in PNAS Nexus put 22 different LLMs and over 100 humans through three standard creativity tests. The results were striking: AI outputs were much more similar to each other than human responses were.

Individual LLMs might outperform individual people. But as a group? The algorithms’ answers clustered tightly together, while human answers scattered across a wide range—9. This pattern held even when researchers explicitly instructed the AI to be more creative—it only slightly increased variability.

A complementary study from HKU Business School, published in Nature Human Behaviour, confirms this pattern. The researchers found that humans displayed huge variation in their creativity scores, ranging from around 35 to 95 out of 100. AI scores, in contrast, fell within a much narrower band of 70 to 85 - 6.

When the researchers randomly sampled 1,000 words from human and AI responses, humans used an average of 577 unique words. AI used only 181 unique words—6.

This is the hidden cost of AI-assisted creativity. It’s not that AI kills creativity—it's that it flattens it. It pulls everyone toward the center. The safe answer. The predictable association. The middle-of-the-road idea that isn’t wrong but isn’t memorable either.

“The problem,” says Yoed Kenett, a cognitive neuroscientist who co-authored the Duke study, “is that while LLMs appear to generate extremely original outputs, they are overly homogenized and not variable in their responses. This could have a detrimental long-term impact on human creative thinking” (9).

Why This Gap Exists: AI Remixes, Humans Originate

The explanation goes to the core of how AI works. Large language models are regression models trained on human data—essentially, the entire public internet. As Professor Haipeng Shen of HKU Business School explains, “They learn from us. On average, they should score the same as us—-6.

AI is a mirror. It reflects patterns it has seen before. It can remix existing ideas brilliantly, but it cannot create something genuinely new—something that hasn’t already been expressed in its training data—2–7.

As India’s Culture and Tourism Minister Gajendra Singh Shekhawat put it at a recent conclave on AI and literature, “AI can describe or summarize content, but it cannot truly experience or understand the emotional depth and inner journey that the writer has expressed. "Originality, imagination, and creative expression remain uniquely human qualities”—2.

This isn’t just philosophy. It’s structural. AI has no lived experience. It doesn’t feel frustration, joy, loss, or wonder. It has never been rejected, heartbroken, or inspired by a sunset. It can describe these things convincingly because it has read millions of descriptions. But it cannot originate them from nothing—7.

Academy of Management scholar Tim Pollock puts it bluntly: “AI has some specific uses in terms of aggregating information… but it’s not a substitute for thinking. And it can’t solve problems, because AI is all about what’s happening now and what it can find on the web that’s been done in the past. It’s not going to help you solve a new conundrum or find new solutions to persistent problems." -7.

The Paradox: AI Lifts the Floor, Lowers the Ceiling

Here’s the tension I keep coming back to. For people who struggle with creative tasks — who face the blank page with genuine anxiety — AI can be transformative. It provides a starting point. It generates options. It lowers the barrier to entry—10.

As the University of Montreal study notes, “Generative AI has above all become an extremely powerful tool in the service of human creativity” (10). It can help average creators become better. It can accelerate prototyping and iteration.

But for people who are already highly creative, the equation changes. As Professor Haipeng Shen warns, “Those already performing well should not rely too heavily on AI because that may diminish their uniqueness” (6).

This is the paradox. AI raises the floor of average creative output while potentially lowering the ceiling of exceptional originality. It makes everyone better at the cost of making the best less distinctive.

The Cultural Consequence: What We’re Losing

This isn’t just an individual problem. It’s a cultural one. The widespread adoption of AI for creative tasks is already reshaping the information ecosystem. According to a May 2026 study by digital marketing agency Graphite, AI-generated English articles surpassed human-written ones in November 2024. As of 2025, over 50% of published articles online are AI-generated, with no sign of decline.

Merriam-Webster named “slop”—the term for low-quality AI-generated content—its 2025 Word of the Year. The New Yorker compared the flood of AI content to the Boston Molasses Flood of 1919: “sticky, suffocating, and slow to clean up” (4).

The deeper danger isn’t just that we’re reading more machine-generated text. It’s that we’re training the next generation of AI on AI-generated data. This creates a feedback loop called “model collapse”—when AI models train on their own outputs, diversity and quality degrade over generations.⁴

As one analysis puts it, “AI writes more; humans write less." Humans write less; AI has less fresh material to learn from. Dwindling material makes AI output more homogeneous. Homogeneous output further reduces the incentive for humans to write”—4".

It’s a vicious cycle. And it’s already spinning.

The Human Edge: What AI Can’t Copy

So where does this leave us? If AI can match average creativity but not exceptional originality, what should humans focus on?

The research points to several distinct human advantages:

Genuine novelty: AI remixes. Humans originate. The most creative humans produce ideas that don’t exist in any training set because they emerge from lived experience, emotional depth, and the messy process of struggling with uncertainty (5–7).

Diversity of perspective: AI models trained on the same internet produce similar outputs. Humans bring wildly different backgrounds, cultural contexts, and personal histories to every problem (6–9).

Embodied knowledge: AI has never touched clay, felt wind on its face, or tasted something unexpected. These sensory experiences feed human creativity in ways AI cannot replicate.

Risk-taking: AI optimizes for probable success based on past patterns. The most creative humans embrace the possibility of failure—and sometimes fail spectacularly before finding something genuinely new.

Meaning-making: AI can generate text. It cannot explain why that text matters, connect it to human values, or make someone feel truly seen. That’s the human face that no algorithm can provide.

The Real Answer: Neither Wins. We collaborate.

After all this research, I’ve stopped asking “who wins.” The question is wrong.

The real answer is collaboration. The HKU researchers developed a platform called “Beat the Bot” where students generate their own ideas, then compare them to AI outputs, then work together in teams to improve both. The result? Higher overall creativity scores and engaged discussion about what makes ideas genuinely original.

This is the model that works. Use AI to generate volume, to escape local maxima, and to explore possibilities you wouldn’t have considered. Then use human judgment to curate, refine, and infuse meaning. The AI provides the raw material. The human provides the soul—10.

As the University of Montreal study concludes, “The future of creativity may lie less in opposition between humans and machines than in new forms of creative collaboration, where AI enriches human ingenuity instead of replacing it” (10).

What This Means for You

If you’re a writer, artist, or anyone who creates, here’s what I’ve learned:

Don’t fear AI. Use it. It’s a tool for generating options, overcoming blocks, and accelerating the boring parts of creation.

But don’t outsource your thinking. The moment you let AI decide what’s good or what’s next, you’ve surrendered the very thing that makes your work valuable—your unique perspective.

Cultivate what AI can’t copy. Your lived experience. Your emotional depth. Your willingness to be weird, to fail, to take risks that don’t optimize for the average -9.

Find your people. The most creative ideas don’t emerge from solo work with AI. They emerge from diverse groups of humans who think differently and aren’t afraid to disagree—5–9.

The research is clear: AI has reached average human creativity. But the most creative among us are still winning. The question isn’t whether AI will replace human creativity. It’s whether you’ll be one of the humans who rises above the average—or one who lets the average become your ceiling.

Thanks for reading. Mubashir