Is AI Actually Creative, or Just Remixing Us?

Ask a diffusion model for a painting or a language model for a short story and the result can be genuinely surprising, good enough that the question almost asks itself: is the machine being creative? The popular answer splits into two camps, both confident and both mostly wrong. One says AI has become a true creative force in its own right. The other says it is hollowing out human creativity and will leave us duller for it. The research suggests something more nuanced than either.

What the machine is really doing

A generative model does not have ideas. It holds a vast statistical map of the text or images it was trained on, and it produces new work by predicting what plausibly comes next. The output can be novel in the narrow sense that no one wrote that exact sentence before, but it stays bounded by the training data in a way human imagination is not. The model recombines. It does not originate, at least not the way a person staring at a blank page does.

This is why AI art so often carries a recognisable house style, and why AI prose tends to drift toward the smooth and the average. It is drawing on the centre of gravity of everything it has seen. That is a real capability, and it is not nothing, but it is a different thing from the leap of a person who connects two ideas no dataset had ever placed side by side.

The creativity it gives, and the kind it quietly takes

The more interesting finding is what happens to us when we use it. A 2024 study in Science Advances by Anil Doshi and Oliver Hauser ran a controlled experiment on short-story writing. Writers who received story ideas from an AI produced work that readers judged more creative and better written, and the effect was strongest for the least experienced writers. So far, so encouraging.

The catch showed up at the group level. Stories written with AI help were markedly more similar to one another than stories written by people working alone. The tool lifted each individual while nudging everyone toward the same middle. The researchers framed it as a social dilemma: each writer is better off, but the collective pool of novel work gets narrower. Individual gain, shared sameness.

Access to generative AI ideas caused stories to be evaluated as more creative and better written, yet those stories were more alike than the ones people wrote unaided.

So where does that leave the myth

The fear that AI will kill human creativity is overblown. Used well, it lowers the barrier to making something, and for a lot of people that is the difference between creating and not. The opposite claim, that AI is now a wellspring of original genius, is just as shaky. It is a superb recombiner working from what we have already made.

The genuine risk is subtler than a robot out-painting Picasso. It is a slow flattening, a world where everyone reaches for the same tool and the same tool reaches for the same middle. The way to avoid that is not to fear the technology but to treat its output as a first draft rather than a final word. The interesting move is still the one the machine cannot make for you. This connects to a point worth keeping in mind: an AI can be perfectly fluent without understanding a word of what it says.