Sometimes I tell people I run an AI art store and they get a face like I just told them I sell AI poetry. The face is fair. The question that goes with it — is AI art real art? — is one I think about every day.
This is my honest answer, written by someone who has stakes in the answer. I’m Emily, I founded Art That Bites, every piece on the site is AI-generated, and my take is more complicated than a yes or no. Below: what AI art actually is, the strongest version of the argument against it, the strongest version for it, where I land, and what I think the question is really asking.
What “AI art” actually means
Most people picture a button. You type “cat in a teacup,” click generate, and a finished image appears. That happens. That’s also the cheapest version of the medium and not what most working AI artists are doing.
The actual workflow looks more like this: a model (Midjourney, DALL-E, Stable Diffusion, Adobe Firefly, take your pick) is trained on millions of images. The artist writes a prompt — sometimes a few sentences, sometimes a few paragraphs, often with style tags, weight modifiers, negative prompts that exclude specific failure modes. The model generates a batch. The artist throws away most of them. They iterate. They feed favored outputs back into the prompt as references. They edit in Photoshop. They composite. Sometimes they regenerate the same image twenty times until one detail finally lands.
What you see on a finished AI art print is the output of dozens of small decisions. The button is real. The artist behind it is also real.
The case AI art isn’t real art
I’m going to make this case as well as I can, because the people making it deserve a good faith argument back.
Training data. The biggest AI image models were trained on enormous datasets scraped from the open web — including, in many cases, the work of professional illustrators and photographers who never consented to it. The models learned what “Greg Rutkowski’s style” or “Studio Ghibli aesthetic” look like by memorizing the work of artists who built those styles over careers. When someone now generates “in the style of Greg Rutkowski,” they’re benefiting from his decades of work without paying him a cent. That’s a real harm. It’s not a vibe. It’s a structural transfer of value from people who made the training data to companies and users who didn’t.
Skill collapse. Traditional illustration is hard. You spend years learning anatomy, composition, color theory, the physics of light. You make a thousand bad drawings before you make one good one. AI flattens that ladder — a person who has never picked up a pencil can prompt their way to something that looks like the work of a trained artist. That isn’t necessarily a moral problem, but it’s a real one for working illustrators, who now compete with output produced in seconds.
Ownership and authorship. If the model did most of the visual work, who is the author? The prompt-writer feels intuitively wrong as the singular answer. Some critics argue AI art is recombinant — statistical pattern-matching dressed up as creation. The output looks like art because it was trained on art, but the system itself doesn’t make art in the way a human artist does. It interpolates.
None of this is wrong. All three points are doing real work in the conversation, and I think anyone selling AI art who pretends otherwise is being dishonest.
The case AI art is real art
Now the other side, also steelmanned.
Photography wasn’t art either, by 1850s standards. When the camera arrived, painters argued it wasn’t a real artistic medium because the machine did the work. The photographer just pointed and pressed. There was a serious, sustained debate in the Royal Academy about whether photography belonged in art exhibitions. They were wrong, but they weren’t stupid. Every new medium gets the same skepticism for the same reasons. The history of art is mostly the history of people getting upset about new tools.
Prompt engineering is itself a craft. Anyone who has actually tried to get exactly the image they imagined out of an AI model knows it’s harder than it looks. The interesting work is in iteration — knowing what to keep, what to throw out, when a result is “almost there” versus “fundamentally wrong,” what tiny prompt adjustment will fix the hand that’s drawn with seven fingers. That skill takes time to develop. People with taste write better prompts. People without taste don’t.
The artist’s job has always been about taste, curation, and intent. Even in traditional media, the artist isn’t doing every brushstroke from scratch — they’re choosing this color over that color, this composition over that one, this final piece out of fifty studies. AI shifts the ratio of execution to curation, but curation has always been a huge part of the work. A photographer takes 200 shots and shows you 4. That’s not less artistic because the camera did the lens work.
The output has a point of view. The strongest defense of AI art isn’t theoretical, it’s empirical: look at what AI artists are actually making. The good ones have voices. You can tell a piece is by a specific artist before you see the credit. That’s not happening accidentally. That’s intent transmitting through the medium.
Where I land
AI art is real art the same way collage was real art and photography was real art — different from what came before, made with different tools, requiring different skills, and inviting the same skepticism every new medium gets when it shows up uninvited.
The complication is that the bottom of the AI art market — the cheap, low-effort, training-data-suspicious end — has been so loud it drowned out the part of the medium that’s actually working. Most of the AI imagery that goes viral is bad. Most of it was made by someone who didn’t iterate, didn’t curate, and doesn’t care. That’s true. It’s also true of every other medium that ever existed. There’s a lot of bad oil painting, bad photography, and bad poetry. The bad versions don’t disqualify the medium.
The training data argument is the one I take most seriously, and it’s the one the medium has to keep working on. Newer models trained on licensed datasets (Adobe Firefly is the cleanest example) are a step toward a more honest version of AI art. Older models trained on scraped work without consent are a real ethical problem, full stop.
If you’re skeptical of AI art, you’re not wrong to be. There’s plenty to be skeptical about. What I’d ask is the same thing I’d ask of someone skeptical of any new medium: judge the work, not the category. Some of it is shallow. Some of it has actual art in it.
What I do differently
I started Art That Bites because I wanted AI art that had a point of view. The market was full of generic dragons and pretty cottages and AI-rendered photorealism that didn’t say anything. I wanted edgy, witty, a little dark. The kind of work you’d hang on your wall because it had something to say, not because it was decorative.
That requires actual artistic decisions. Each design starts as a prompt — usually one I’ve been refining for hours or days — gets generated, gets iterated, gets thrown away or kept. Most of the prompts I write don’t make it to the shop. The ones that do are the ones with a voice. Punk princesses didn’t happen in one click. Steampunk pineapples took a stupid amount of iteration before they earned their absurdity.
I’m also made-to-order on Printful, which means there’s no warehouse of unsold AI prints sitting around taking up space and waiting to become landfill. Every piece is printed when someone buys it. Less waste. Less throwaway content. The medium is still new enough that how the prints get made is part of the answer to whether they’re real art.
The question your customers are actually asking
“Is AI art real art” is rarely the actual question. The actual question is one of three things:
- “Will I feel weird hanging this on my wall?” The honest answer: only if you wouldn’t feel weird hanging anything on your wall. The piece on your wall does the work the piece on your wall does. It doesn’t ask you to defend it at parties.
- “Did this take any actual effort?” Yes. The good ones do. The bad ones don’t. Same as anything else.
- “Am I supporting something ethical?” That depends on the artist. Look at what model they use, where the design came from, whether the work has a real point of view. The market is mixed. Some of it is fine. Some of it isn’t.
You bought the print. Did you buy art? Yes, by any reasonable definition of art, you bought art. The thing on your wall was made for a reason, with a point of view, on purpose. That’s the test. It always was.
Frequently Asked Questions
Are Art That Bites prints actually AI-generated?
Yes. Each design starts as a prompt, gets iterated through dozens of generations, and only the strongest get sold. We don’t pretend otherwise — we’ve also never tried to pass the work off as anything other than what it is.
Do AI art companies pay royalties to artists in the training data?
It depends on the model. Newer commercial models — Adobe Firefly is the most notable — are trained on licensed content where the contributing artists were paid or opted in. Open-source models like older Stable Diffusion versions weren’t. The industry is moving toward licensed training data, slowly, and pressure from artists is what’s moving it.
Will AI replace human artists?
It’s already changing what “artist” means, the same way photography did to portrait painters in the 1860s. Photographers didn’t replace painters; they changed what painting was for. The painters who survived figured out what photography couldn’t do and went there. The same dynamic is happening now, faster.
How do I tell good AI art from bad?
Same way you tell good non-AI art from bad. Does it have a point of view? Did the artist make decisions, or did the model? Does the piece say something specific, or could it be any random output of any random prompt? Look for intent, taste, and bite. The bad AI art is generic. The good AI art has voice.
Should I buy AI art prints?
If a piece moves you, yes. If it doesn’t, no. The medium isn’t a moral category — it’s a technique. The question to ask is the same one you’d ask of any print: do you want this on your wall every day for the next ten years? If yes, it’s worth buying. If no, it isn’t.
If you got this far
Browse the Art That Bites collection. Or read more on the blog: Dark Academia Wall Art if you came here looking for an aesthetic, or the Wall Art Size Guide if you came here looking for practical help. Either is a less philosophical use of your time. Both will hang on a wall just fine. Get Smitten.