The short answer is yes – if creativity means producing something new, surprising, and useful. But if creativity also requires a point of view, emotional experience, intention, and something personally at stake, today’s AI is not creative in the way a person is.
AI can make an artistic product. It cannot yet live the human process that gives art its meaning.
That distinction matters. It allows us to take AI’s abilities seriously without confusing fluent generation with imagination, or treating a powerful creative tool as an artist with an inner life.
How does AI make art?
Generative AI learns patterns from enormous collections of human-made material. A language model learns relationships among tokens in text. Image, audio, and video systems encode their inputs into mathematical representations and learn how features tend to relate to one another.
It is tempting to say that AI converts everything into language. That is close as a metaphor, but not technically exact. Some multimodal systems do place text, images, and video into a shared sequence of tokens; others work in a latent mathematical space. In either case, the machine does not see a sunset, hear a minor chord, or feel the weight of a sentence as a person does. It processes representations and predicts what should come next.
From those learned relationships, it can generate a new combination: a portrait with the geometry of Art Deco, a melody that carries the structure of a lullaby, or a paragraph shaped by thousands of examples of persuasive writing. A recent Nature paper on multimodal learning describes one version of this process in direct terms: images, text, and video are tokenized and generated through next-token prediction.
That is an astonishing technical achievement. But prediction is not experience.
Is AI creating, or just remixing?
“AI only remixes what already exists” sounds decisive, but it is not the whole argument. Human artists also work with inherited forms, remembered images, cultural traditions, and other people’s influence. No novelist invents language from nothing (JRR Tolkien did, but he’s a different level.). No musician composes without a history of rhythm, harmony, or sound.
Novel combinations can be creative. The harder question is whether a novel output alone is enough to make its producer a creator.
On narrow tests, AI performs impressively. A 2023 study in Scientific Reports compared 256 people with three AI chatbots on an alternate-uses task, in which participants imagined unusual uses for everyday objects. The chatbots scored above the human average, although the best human responses still matched or exceeded the AI’s ideas.
Other researchers have proposed “relative” or “statistical” creativity: instead of asking whether a machine possesses creativity in some universal sense, ask whether its outputs are indistinguishable from those of a particular group of human creators. Under that definition, the authors of Can AI Be as Creative as Humans? argue that an AI could theoretically match the creative ability represented in the human data it learns from.
These findings tell us something important: AI can produce outputs that people judge to be creative. They do not prove that the machine understands what it made, wanted to make it, or experienced the emotion that the work appears to express.
Why emotion and intention still matter
Imagine an AI-generated song about grief. It may contain the fragile melody, unresolved harmony, and intimate language that people associate with loss. A listener may cry. The emotional response is real but it belongs to the listener.
The system itself has not lost anyone.
This is where human and machine creativity separate. Human art is not only a pattern in the final product. It is an act performed by someone with a body, a history, relationships, memories, fears, desires, and a reason to speak. Even highly conceptual or deliberately unemotional art carries intention: a person chose the form, accepted the risk, and placed the work in a social world.
Current AI can model the expression of sadness without being sad. It can describe love without loving and imitate urgency without wanting anything to change. As creativity researcher Mark Runco argues in a BrainFacts discussion of AI and creativity, intentionality is a central part of human creativity. A model responds to a prompt; it does not decide that something inside it needs to be expressed.
This does not make AI-generated work worthless. Art has always depended partly on what an audience brings to it. A machine-made image can provoke a human idea, reveal an unexpected association, or become raw material in a larger human work. But the meaning comes from the people who prompt, select, reject, edit, frame, and encounter it.
AI can participate in the making of art without becoming the emotional subject of that art.
The hidden trade-off: Stronger ideas, less variety
AI’s greatest creative strength may also be its greatest limitation. It has learned from a huge portion of the shared cultural record, so it can quickly produce a polished answer near the center of what tends to work. When many people ask similar questions, however, they can be guided toward the same center.
Research highlighted by Wharton found that generative AI could improve the quality of an individual’s ideas while reducing diversity across a group. In one experiment, only 6% of AI-assisted ideas were judged unique, compared with 100% in the human-only group. Across 45 comparisons, AI-assisted ideas were significantly less diverse in 37.
That is the difference between producing one strong idea and sustaining a creative culture. Innovation depends on variation: unusual memories, local knowledge, disagreement, mistakes, obsessions, and perspectives that do not come from one shared system.
A 2026 Georgetown University report on human and AI creativity makes a related point. Researchers examining more than 370,000 college admissions essays found that essays written after ChatGPT’s release used a wider range of words but often contained fewer original ideas. Sophisticated language can create the appearance of creativity while masking convergence underneath.
If every writer, designer, and strategist begins from the same generator, creative work may become more polished and less distinct at the same time.
Why creative people may benefit most from AI
The people most capable of using AI creatively are often those who already have taste, curiosity, and a point of view. AI can lower the cost of trying an idea, but it cannot decide which idea deserves to exist.
A designer can generate ten rough compositions before drawing one. A musician can test arrangements without booking a studio. A writer can expose gaps in an argument, explore alternative structures, or edit a difficult passage. A filmmaker can build a storyboard before raising money for a shoot. In each case, AI removes friction between imagination and experiment.
That does not mean adding AI automatically improves the result. In three studies involving human–AI teams, researchers found that creativity improved through deliberate back-and-forth refinement – not through endlessly asking the model for more ideas. As the University of Cambridge’s summary of the research explains, the strongest results came when people and AI developed ideas together with clear guidance and feedback.
The useful model is not “AI makes the art.” It is:
- A person begins with an observation, question, feeling, or rough idea.
- AI helps expand possibilities, test versions, or handle repetitive work.
- The person challenges the obvious options and makes the meaningful choices.
- The finished work remains grounded in human judgment and responsibility.
Used this way, AI does not replace creativity. It gives a creative person a faster sketchbook, a tireless sparring partner, and access to forms of execution that may once have been too expensive or technical.
The danger begins when convenience moves AI to the first and last step—when the person contributes neither the original impulse nor the final judgment.
Govern the data, protect the human contribution
The most urgent questions about AI art may not be philosophical. They are practical: What material was used to train the model? Did creators consent? Can they opt out? Should they be credited or paid? Can audiences tell when media has been generated or manipulated? Who is responsible when a system imitates a living artist or reproduces protected work?
Responsible governance should focus on these concrete points of power: training-data transparency, permission and licensing, compensation, provenance, labeling, privacy, and accountability. In the European Union, the current framework for general-purpose AI already emphasizes transparency about training inputs and compliance with copyright obligations. These are more useful safeguards than trying to make a regulator decide whether an image contains a human “soul”.
Law also reflects the continuing importance of human authorship. The U.S. Copyright Office’s guidance says that using AI as an assistive tool does not prevent a work from receiving copyright protection. But purely AI-generated material is protected only when a human has contributed sufficient expressive authorship; prompting alone is generally not enough.
The principle is sensible: regulate the industrial system around AI without regulating away human experimentation. Protect the people whose work feeds the models. Preserve attribution and choice. Require transparency where synthetic media could mislead. And leave room for artists to use new tools in ways no policy-maker—or model developer—can predict.
So, can AI make art?
AI can generate images, music, stories, and films that satisfy many practical definitions of art. It can surprise us. It can outperform the average person on specific creativity tests. In the hands of a skilled creator, it can make ambitious work possible.
But current AI has no lived experience behind its output. It has no private memory it is trying to preserve, no grief it needs to transform, no community it hopes to challenge, and no future it wants to change. It can reproduce the language of emotion across words, pixels, and sound, but it does not feel what that language means.
For now, people still provide the source of art: the impulse, the intention, the taste, the risk, and the emotional stakes. AI can help us make more. It may even help us make better. But it should expand human expression, not become an excuse to stop practicing it.
The real threat is not that machines will suddenly develop a soul. It is that people, seduced by speed and polish, will stop putting their own into the work.
Frequently Asked Questions
Can AI create something original?
AI can produce a combination that has never appeared before, so its output can be original in a practical sense. However, it generates from patterns learned in existing data and does not originate work from lived experience or independent intention.
Will AI replace artists?
AI will automate parts of creative production and change some creative jobs. It is less likely to replace the human qualities that make work culturally meaningful: taste, intention, context, emotional experience, and responsibility. Artists who learn to direct and edit AI may gain new capabilities.
How can artists use AI without losing their creativity?
Start with your own idea before consulting AI. Use the model for options, critique, iteration, or repetitive execution. Then reject generic outputs, add specific human experience, and make the final decisions yourself. AI is most useful as part of a creative process, not as a substitute for one.
