The Impossible Standard We Suddenly Expect from AI
One thing fascinates me.
When a human photographer makes something reminiscent of Peter Lindbergh, Steven Meisel or Glen Luchford, we call it influence.
When AI makes something reminiscent of existing fashion imagery, people suddenly expect the machine to invent visual language from nothing.
As if creativity has ever worked that way.
Every creative system learns from previous work.
Humans do too.
The difference is that human learning feels invisible. It happens slowly, over years, through magazines, films, museums, campaigns, dreams, mistakes and obsessions.
AI learning feels uncomfortable because it happens all at once.
And because it makes the process harder to romanticise.
⸻
The Filippa K Case
The Filippa K AI campaign is a useful example because the criticism was not vague.
Two specific references were pointed out: one image was said to closely resemble work by Athina Chrissaki, and another was said to resemble a photograph by Johan Sandberg, shot for ARKET’s 2021 collection. Swedish media reported that the campaign was paused after accusations that two AI-generated images looked suspiciously similar to existing photographs by those photographers.
That matters.
Because this is not only a conversation about AI “looking like fashion.”
It is a conversation about what happens when an image lands too close to a specific, traceable source.
And that is where the debate becomes serious.
⸻
Is the Pose Even Unique?
Let’s be honest.
How many ways are there to photograph a man pulling off a sweater on a windy beach, against blue sky, in natural light, for a Scandinavian fashion brand?
Not infinitely many.
Some poses repeat because they work.
They show movement.
They show texture.
They show silhouette.
They sell the garment.
If we searched fashion archives from the last twenty years — Vogue editorials, Calvin Klein, COS, Zara, Massimo Dutti, ARKET — we would almost certainly find dozens of images that feel related.
Not because everyone copied each other.
Because fashion has a visual grammar.
And that grammar repeats.
⸻
Fashion Has Always Had Copycats
This is the part people sometimes pretend not to know.
Fashion has always lived in the grey zone between homage, reference, theft, archive, moodboard and repetition.
There is a reason accounts like Diet Prada became powerful. Their entire format was built on placing new fashion images next to older, suspiciously similar work and asking the question the industry hates:
Is this inspiration, or is this copying?
And this is not limited to clothes.
Images copy images too.
Madonna was sued by the son of fashion photographer Guy Bourdin, who claimed her Hollywood video imitated Bourdin’s work.
Jeff Koons was accused of plagiarising a fashion advertisement for Naf Naf.
Gucci faced plagiarism accusations when a campaign video was compared to the work of a Central Saint Martins student — a case Gucci denied, arguing that both drew from the same 1970s science-fiction references.
H&M entered a copyright dispute after using graffiti artist Revok’s mural in a campaign without permission.
So no, visual borrowing did not begin with AI.
AI has simply made it faster, more visible, and harder to excuse.
⸻
We Rarely Ask Humans the Same Question
Imagine two photographers independently photograph a man removing a knit sweater on a beach.
Would anyone immediately accuse one of copying?
Probably not.
Now imagine both photographers had spent twenty years studying the same magazines, the same campaigns, the same art books, the same cultural references.
Of course their images would overlap.
Nobody expects every human photograph to emerge from a vacuum.
So why do we suddenly demand that AI create as if nothing came before it?
⸻
The Double Standard
Creativity is rarely pure invention.
It is usually recombination.
Painters borrow compositions.
Directors borrow camera language.
Photographers borrow poses.
Fashion borrows entire decades.
Sometimes intentionally.
Sometimes subconsciously.
The history of art is largely the history of reinterpretation.
The difference is that when humans do it, we often call it taste.
When AI does it, we call it theft.
That double standard is too simple.
And too comfortable.
⸻
Similar Doesn’t Always Mean Copied
That does not mean every AI output is acceptable.
If a system reproduces a protected image so closely that it becomes effectively the same work, that is a real concern.
The Filippa K debate shows exactly why that line matters.
There is a difference between an image that belongs to a broad visual tradition and an image that appears to mirror a specific photograph too precisely.
But similarity exists on a spectrum.
The real question is not:
“Does this resemble another image?”
Everything resembles something.
The real question is:
“At what point does influence become reproduction?”
That question existed long before AI.
AI has simply forced us to confront it more honestly.
Image note: The images in this post are AI-generated studies created as part of this essay. They are not affiliated with Filippa K, ARKET, Athina Chrissaki or Johan Sandberg, and they are not presented as original campaign images. They exist here as visual arguments: an exploration of reference, repetition, transformation and the uneasy line between influence and imitation. 
Post a Comment