Can AI Be Creative?

It depends entirely on how you define creativity. AI can produce genuinely novel outputs that surprise even its creators — but it lacks the intention, emotion, lived experience, and self-awareness that most definitions of human creativity require.

When a Midjourney image wins a state fair art competition or an AI-generated novel gets shortlisted for a literary prize, it forces a real question: is that creativity? The answer reveals as much about our definition of creativity as it does about AI. This article walks through the strongest frameworks for thinking about machine creativity, what the research actually shows, and where the real debate lies.

How Researchers Define Creativity

The most rigorous framework for evaluating machine creativity comes from cognitive scientist Margaret Boden, who distinguished three types in her 2004 book The Creative Mind. Combinational creativity produces new combinations of familiar ideas — like mixing jazz with classical. Exploratory creativity stretches the rules of an established conceptual space — like Picasso distorting perspective within painting. Transformational creativity breaks the rules entirely, creating a new conceptual space — like Einstein redefining time.

By Boden’s framework, current AI clearly achieves combinational creativity and arguably reaches exploratory creativity in some domains. Transformational creativity — genuinely inventing new conceptual frameworks — has not been demonstrated. Every AI “creative” output is a recombination of patterns present in training data, however sophisticated that recombination becomes.

What AI-Generated Art Actually Demonstrates

In September 2022, Jason Allen’s Midjourney-generated image Théâtre D’Opéra Spatial won first place in the digital art category at the Colorado State Fair. This triggered a fierce debate about whether AI art was “real” art. The painting community was outraged; AI advocates pointed out that Allen had spent 80+ hours refining prompts and selecting outputs.

In 2023, an AI-assisted artwork sold at Christie’s New York auction for $1.08 million, setting a new benchmark for AI-generated creative work’s market value. These are not flukes — they reflect genuine aesthetic quality that trained human judges found compelling. Whatever you call it, the output achieved the social function of art: it moved people and commanded market value.

Tools like DALL-E 3, Midjourney v6, and Open Art AI can now generate images that are technically indistinguishable from professional illustrations in many contexts. Adobe Firefly takes a different approach — trained exclusively on licensed and public domain content — making it a legal alternative for commercial use. The quality gap between AI and human commercial illustration has effectively closed for standard work. The gap at the highest levels of intentional human expression remains wide.

The Intention Problem

The strongest argument against calling AI outputs “creative” is the absence of intention. When Picasso fragmented human faces in his cubist period, he intended to destabilize bourgeois certainties, challenge the viewer’s assumptions, and create a new visual language for 20th-century experience. The meaning was inseparable from the intent. AI has no such intent. It produces outputs that statistically match a prompt, optimized against human preference feedback. The “creativity” is in the prompt writer, not the model.

This is not a semantic technicality. It changes how we evaluate the work. A human artist who produces derivative, predictable work is doing less than their potential. An AI that produces the same derivative work is doing exactly what it was built to do. The frame of evaluation is completely different. When you use AI image generators for professional work, you are the creative agent; the AI is a highly sophisticated brush.

What AI Lacks: Emotion, Stake, and Surprise

Human creativity often emerges from emotional necessity. Frida Kahlo painted her pain. Beethoven composed through deafness. The stakes — personal, existential, social — shape what gets made and why. AI has no such stakes. It cannot be desperate, delighted, heartbroken, or bored. It does not “want” to make anything. This absence is not just philosophical — it shows up in the work. AI creative outputs trend toward the aesthetically pleasing and technically competent, but rarely toward the raw, uncomfortable, or genuinely strange outputs that human artists produce when they’re working from deep necessity.

A 2023 study from Georgia Tech (arXiv:2311.07071) tested whether human evaluators could detect “creative surprise” — outputs that violated expected patterns in meaningful ways — in AI versus human creative writing. Human writers produced surprising-but-coherent outputs significantly more often. AI outputs were more technically consistent but less surprising. The researchers concluded that current models optimize for plausibility at the expense of genuine creative risk.

Where AI Creativity Is Genuinely Valuable

None of this means AI is useless for creative work — quite the opposite. AI excels at the early-stage, generative parts of creative processes: ideation, variation, style exploration, and removing the blank-page problem. Writers use AI to generate first-draft options they then shape and refine. Designers use it to explore 50 visual directions in an hour rather than 5 in a day. Musicians use it to generate melodic fragments they then develop. The creative intelligence is the human’s; the AI dramatically expands the option space they can explore.

The most productive frame is AI as a creative collaborator with extraordinary pattern knowledge and zero creative judgment. It knows what a sonnet looks like, what noir cinematography looks like, what a brand identity in the style of Swiss modernism looks like — it has processed millions of examples of each. It can show you versions at speed. Whether those versions become meaningful art depends entirely on the human in the loop. Check out our guide on AI writing tools to see how writers are integrating these capabilities practically.

The Copyright Question

US copyright law currently holds that AI-generated works with no human creative input are not copyrightable. The Copyright Office’s 2023 guidance established that “works produced by machine or mere mechanical process” without human authorship cannot be protected. This has real commercial implications: if you use AI to generate images or text for commercial products, the output is in the public domain unless you can demonstrate sufficient human creative contribution. This is a rapidly evolving area — see our detailed explainer on AI and artists’ copyright for the latest legal landscape.

---

Key Takeaways

• Boden’s framework distinguishes combinational, exploratory, and transformational creativity — AI clearly achieves the first, possibly the second, and has not demonstrated the third.
• AI-generated work can achieve real aesthetic quality and market value, as the 2022 Colorado State Fair and 2023 Christie’s sale demonstrated.
• The absence of intention, emotion, and stake distinguishes AI output from human creative expression in important ways.
• AI is most valuable as a creative accelerator — expanding the option space for human creative decision-making.
• US copyright law does not protect AI-generated work without meaningful human creative contribution.

---

Frequently Asked Questions

Can AI write a novel that’s actually good?

AI can write fluent, technically competent long-form prose. Whether it is “good” depends on what you value in fiction. AI output tends to be coherent and well-structured but lacks the idiosyncratic voice, emotional depth, and risk-taking that distinguish literary fiction. Several AI-assisted novels have reached shortlists; none have won major literary prizes on their own.

Is using AI tools cheating in creative work?

This depends entirely on context and community norms. In fine art competitions, most now require disclosure of AI involvement. In commercial writing, clients increasingly specify whether AI assistance is permitted. In personal creative practice, using AI is no different from using Photoshop or spell-check — it’s a tool. The ethical question is always about disclosure and context.

Will AI replace human artists?

For commodity creative work — stock illustrations, background music, template copy — AI has already displaced significant human labor. For high-end, intentional, emotionally resonant creative work with genuine authorial voice, human artists remain irreplaceable. The middle tier — competent commercial creative work — is under the most pressure.

What is the best AI tool for creative writing?

For long-form creative writing, Claude 3.5 Sonnet and GPT-4o are the current leaders in maintaining narrative coherence. For poetry and stylistic experimentation, Claude tends to take more interesting creative risks. For pure volume production — blog posts, marketing copy — any modern model works well. See our AI writing tools comparison for current pricing and feature details.

Does AI-generated music count as music?

By any functional definition, yes. Suno and Udio generate full songs with vocals, arrangement, and production that are indistinguishable from human-produced music in casual listening. Whether it qualifies as “art” is the same philosophical question as AI visual art — and has the same answer: it depends on your definition of art, not on the sound itself.

---

If you are new to AI and want to understand the technical foundations behind these creative capabilities, our primer on what artificial intelligence is gives the grounding needed to evaluate AI art and writing tools in context.

Sources: Boden, M. (2004). The Creative Mind: Myths and Mechanisms; Georgia Tech arXiv:2311.07071; US Copyright Office, “Copyright and Artificial Intelligence” (2023); Grokipedia: Generative AI Art; Christie’s auction records, 2023.

You May Also Like

• What Is Artificial Intelligence
• Best AI Tools for Beginners
• How to Use AI
• AI Tools Directory
• Best Free AI Courses