AI can now generate paintings, music, stories, films, and ideas that can surprise even its creators. But producing something new isn’t necessarily the same as imagining it.
As machines become increasingly capable of creative work, the uncomfortable question is no longer whether AI can make art. It’s whether a machine can imagine.
Introduction: The uncomfortable question
Imagine a museum in 2040.
One gallery displays paintings created by humans. Another displays paintings created entirely by machines.
Visitors can’t tell which is which.
Then comes the unsettling part: a machine-generated painting wins the exhibition’s highest artistic award.
Would we call the machine an artist?
This isn’t science fiction anymore. Recent research shows that generative AI can perform remarkably well on conventional measures of creativity. A 2025 study found several leading AI models outperforming human participants on divergent and convergent thinking tasks (Arora et al., 2025).
A much larger 2026 study comparing more than 9,000 humans with LLMs found that humans retained an advantage in average creativity, while some AI systems demonstrated striking performance at the high end of creative idea generation (Wang et al., 2026).
So, yes, machines can produce surprisingly creative outputs.
But is that imagination?
AI can generate, but can it imagine?
Here’s where things get uncomfortable.
AI can generate an image of a city floating inside a human eye, invent an imaginary creature, or write a story about a civilization that exists on another planet.
It can produce combinations that didn’t previously exist.
But human imagination isn’t simply combination.
We imagine through memories, emotions, bodily experiences, fears, desires, relationships, culture, and personal history.
A 2025 art-science analysis argues that human creativity is deeply connected to embodiment, perception, emotion, material interaction, and the ability to identify problems — not merely generate patterns (Candy et al., 2025).
That’s a serious distinction.
The creativity paradox
AI may produce something original without having an inner experience of originality.
That sounds contradictory, but recent research makes the distinction important. AI models can score extremely well on creativity tests, yet those tests may measure only particular forms of creative performance rather than imagination as a broader human capacity (Arora et al., 2025).
There’s another problem.
Research published in Nature Human Behaviour found that ChatGPT can improve the creativity of individual ideas while reducing the diversity of ideas produced collectively (Meincke et al., 2025).
A 2025 study similarly reported a homogenizing effect: repeated AI-generated writing contributed fewer new ideas than repeated human writing (Zhou et al., 2025).
Imagine a future where billions of people have the same AI creative assistant.
We might get millions of beautiful ideas — and fewer genuinely different ones.
The machine artist
Research on human-AI co-creativity is already moving beyond the simple question of whether AI is creative.
Medeiros (2025) found no consistent evidence that exposure to AI-generated creative material automatically improves human divergent thinking, suggesting that AI assistance isn’t a guaranteed creativity booster.
Other 2026 research suggests that the quality and nature of human-AI interaction matter: different forms of interaction can influence novelty and usefulness differently (Yuan & Peng, 2026).
So perhaps the future artist won’t be human or machine.
Perhaps it’ll be both.
Conclusion: The question we should be asking
By 2040, machines may create paintings more technically impressive than most humans can produce, compose music that moves millions, and write stories that make us cry.
But technical capability isn’t necessarily imagination.
AI may generate possibilities. Humans give those possibilities intention, meaning, experience, and context.
And perhaps that’s the real controversy.
The machine doesn’t need to become human to transform creativity.
It only needs to become good enough to make us question what creative means being.
Who knows? In the far future, the machine may not just imitate imagination — it may develop a form of its own.
