Why the same work is brilliant if it's human and rubbish if it's AI

The experiment that should be taught in schools

Here is an experiment that researchers have run several times since 2020, with variations but always the same result. You take an image, a text, or a piece of music. You show the same work to two groups of people. To the first group, you say: "made by a human". To the second: "made by an AI". Then you ask each person what they think of it.

The results are brutal and reproducible. According to a study published in Scientific Reports (the scientific journal of the Nature group) in 2023, simply labelling a work as AI-generated causes a 62% drop in perceived monetary value and a 77% drop in estimated production time. For the same work. Strictly identical. Only the label changes.

Another study published on arxiv (the pre-print platform used by researchers before official peer review) in 2024 goes further. When evaluators are asked to compare two texts blind, they cannot tell the human from the AI. But as soon as their origin is revealed (or claimed to be revealed), they prefer the work labelled "human" by 30%+. And here is the detail that should make everyone think: the preference persists even when the labels have been deliberately swapped. That is, people prefer what they are told is human, even when it is in fact AI.

This article is an honest exploration of that collective hypocrisy. Not out of naivety. Not to defend AI. But to look squarely at a massive psychological phenomenon that shapes our relationship with technology. And which, above all, says something deeply unflattering about us.

The documented paradox: we like AI when we don't know

Here is the detail that sums it all up. A study published in Nature Communications (another major scientific journal of the Nature group) in 2025 showed that in a blind comparison, where participants do not know which work is human and which is AI, they slightly prefer the AI works. Better composition, more balanced colour harmony, cleaner finishing. The image-generation models of 2024-2025 produce works that are technically superior to the average of professional human output.

Except that as soon as the origin is revealed or claimed to be revealed, the judgement flips. The same work, which was aesthetically preferred 30 seconds earlier, suddenly becomes "cold", "empty", "soulless", "not creative", "doesn't deserve to be called art". The same brain, looking at the same image, produces two opposite judgements based on contextual information that changes nothing about what is actually seen.

If we took this mechanism seriously and applied it to any other field, we would not hesitate to call it a major cognitive bias. If someone systematically preferred the same food when told it was French rather than Belgian, or that it was made by a woman rather than a man, we would call that chauvinism or misogyny. When it comes to AI, we call it "keeping your humanity". It changes nothing about the underlying psychological mechanism.

A concrete case: Jason Allen, the perfect culprit

In August 2022, Jason Allen, a 39-year-old American game designer, won the "Digital Arts / Digitally-Manipulated Photography" category at the Colorado State Fair (the annual state fair of Colorado, one of the oldest in the United States) with a work entitled "Théâtre d'opéra spatial". The work is beautiful. Baroque composition, deep colours, dramatic atmosphere. The jury awarded it first prize.

Then Jason Allen did something he need not have done: he revealed that the work had been generated with Midjourney (one of the leading AI image-generation services, accessible via Discord). He had stated this on the entry form. He hid nothing. He was simply transparent.

The reaction was a deluge. Dozens of indignant articles. Thousands of furious tweets. Professional artists crying foul, crying the death of art, crying the end of human creation. The jury upheld the prize, but the controversy was massive. Jason Allen received death threats. Several commentators called for the rules to be changed to ban AI works from any competition in future.

Here is the question none of those indignant voices asked. If the jury judged the work beautiful enough to give it first prize without knowing its origin, does that beauty disappear retroactively because we know its origin? The work is exactly the same. It was beautiful enough to win a competition when its origin was unknown. Why does it become unworthy as soon as we know how it was produced?

The honest answer is that the quality of the work has nothing to do with the anger. The anger comes from something else. From perceived economic competition. From a challenge to the identity of the artist. From a fear of replacement. From a fear that what was mysterious and difficult has become easy. None of that anger concerns the work itself.

The everyday hypocrisy we prefer not to see

Beyond the Jason Allen case, the collective hypocrisy is measurable in daily life. Here are a few observations.

Observation 1: we use AI on the sly. According to the 2025-2026 statistics we documented in our article on AI in education, 88% of students now use ChatGPT for their assessments. But almost none of them admit it. The same person who publicly slags off "AI art" quietly asks ChatGPT to draft their monthly report, their email to a difficult client, their covering letter. Public contempt, private daily use.

Observation 2: we love it when a famous human uses AI. Bruno Mars releasing a track with AI-enhanced vocals: applause. An unknown artist releasing an AI song on Spotify: "it's shit". Spielberg using AI effects in his next film: technological avant-garde. A YouTuber doing the same: job thief. Hayao Miyazaki, Oscar winner, who has reportedly used a bit of AI for his next Ghibli: we'd call that "creative integration". A 19-year-old doing it: "destroying animators' livelihoods". The same act, two opposite judgements depending on the status of the person doing it.

Observation 3: we're blind to AI when it suits us. You're listening to Spotify. A growing share of the ambient tracks you hear ("chill lofi for studying", "morning coffee jazz", "rainy day playlist") is now AI-generated. You don't know, and it doesn't bother you. You look at a Netflix poster thumbnail: AI-generated. No problem. You read an automatic product description on Amazon: AI-generated. No scandal. AI is everywhere, setting the mood of your life, and you don't level a single criticism at it. Until someone warns you. And then, suddenly, it's unacceptable.

Observation 4: we forgive human flaws, never AI's. When a human journalist makes a factual error in an article, we call it "a typo" or "an approximation". When an AI does the same, we call it a "hallucination" and conclude that AIs are fundamentally unreliable. When a human painter gets a perspective wrong, it's a "style". When an AI gets a hand wrong with 6 fingers, it's "proof that AI understands nothing". The same kind of flaw is interpreted completely differently depending on its origin.

The "it's not real art" rejection: a deconstruction

The most common argument against AI is: "it's not real art, it's not real creation, it's not real writing". Let's examine it honestly.

"There was no human effort." False. Someone wrote the prompt, selected from dozens of generations, reworked the image, made aesthetic choices. The effort is different from that of a traditional painter, but it exists. If we reject AI works because "there was no effort", we must be consistent: also reject photography (just a click), the ready-made (Marcel Duchamp and his signed urinal in 1917), conceptual art (Sol LeWitt, who had his works made by others), a good deal of contemporary art.

"There is no creative intention." Debatable. The prompt is an intention. The choice between ten generations is another. The retouching is a third. In the end, the work carries a human's choices, just with a different technical mediation. Just as photography carried the photographer's intention despite the mediation of the camera.

"AI has no sensibility." True. But neither does the brush, the sewing machine, the camera, the Photoshop software. No tool has sensibility. The sensibility is the user's. When you look at a photograph by Sebastião Salgado (a Brazilian photographer famous for his documentary images), you see Salgado's sensibility, not his Leica's.

"It's cheating." Here the argument becomes more precise. But then the question is: cheating compared with what? If you enter a competition that requires a manually made work, and you use AI in secret, then yes, it's cheating. If you sell a work as "hand-painted" when it is partly AI, that's fraud. But using AI when it is permitted and disclosed is not cheating: it is simply using a different tool. The pencil was "cheating" compared with prehistoric charcoal. The typewriter was cheating compared with the quill. The computer was cheating compared with the typewriter. At every stage, the old guard cries cheating. At every stage, history proves them wrong.

"AI exploits artists' work without paying them." Here we come to an economically and legally legitimate argument, one that deserves a real discussion. As we explored in our article on journalism, the training of AI on protected works without consent or compensation is a real problem, and one that is the subject of ongoing lawsuits. But that argument is not an aesthetic judgement on the work produced. It is an economic argument about the mode of production. Confusing the two is precisely the hypocrisy we are denouncing here.

The double standard with previous tools

To put the hypocrisy in perspective, let us look at what humans gradually accepted as tools in art, without crying the end of civilisation.

The camera (1839). When photography appeared, several painters called it "the death of art". Charles Baudelaire (French poet and art critic, author of The Flowers of Evil) wrote in 1859: "Industry, by bursting into art, becomes its deadliest enemy." 200 years later, nobody seriously disputes that photography is an art form in its own right, and several photographers (Henri Cartier-Bresson, Annie Leibovitz, Sebastião Salgado) are regarded as great artists.

Sampling (the 1980s). When hip-hop began taking excerpts from other tracks to make new music, classical critics cried theft, the end of original music, creative laziness. Today, sampling is regarded as an art form in its own right, and several producers (DJ Premier, J Dilla, Kanye West) are studied in conservatoires.

Photoshop (the 1990s). When Photoshop arrived, you heard that "it's not real photography any more", that "everything is faked", that "anyone can do it now". Thirty years later, Photoshop is a standard professional tool, used by 90% of pro photographers, and nobody disputes the art of the great photographers who use it extensively.

Auto-Tune (the 2000s). When T-Pain popularised Auto-Tune (software that corrects vocal pitch in real time) in hip-hop, people said "these singers are frauds, they can't sing". Today, Auto-Tune is part of the standard artistic vocabulary of pop and hip-hop, and several artists (Travis Scott, Lil Uzi Vert) make recognised creative use of it.

In all these cases, the initial reaction was rejection, mockery, an accusation of cheating. In all these cases, the technology ended up being integrated and accepted. There is no serious reason to think that AI will follow a different trajectory, unless we assume that people today are more intelligent than all their predecessors. Which would be, statistically, unlikely.

Why we are hypocrites: the underlying psychology

To understand the scale of the phenomenon, we need to name the real psychological reasons for this bias. They are multiple and they compound one another.

Anthropocentrism. Humans love to see themselves as special, unique creators, irreplaceable. Any technology that suggests otherwise is psychologically threatening. This is what psychologists call the "anthropocentric creativity bias": we assume a priori that creativity is intrinsically human, and therefore that a non-human creation cannot be real creativity, by definition. It is circular logic, but it is powerful.

Economic fear. Part of the rejection comes from people who are economically threatened by AI: illustrators, writers, musicians, designers. Their rejection is not aesthetic, it is protectionist. That is entirely understandable on a human level. But it is important to name it for what it is: an economic argument dressed up as a moral one.

Effort as a value in itself. Christianity, and more broadly modern Western culture, treats effort as a virtue (the "work ethic"). If something looks easy, it is suspect. AI makes creation look easy, so it is suspect. This intuition has no aesthetic basis: a work is not more beautiful because it cost more suffering. But it carries deep moral weight in our culture.

Identification with the creator. When we look at a human work, we identify with the creator. "I could be the one who painted that." When we look at an AI work, that identification disappears: we are not an AI. This absence of identification reduces aesthetic pleasure. It is measurable, it is documented, and it is entirely understandable. But it is also a reason that says more about our narcissistic need for identification than about the actual quality of the work.

Herd effect. In 2025-2026, disparaging AI is socially rewarded in certain cultural circles (artists, intellectuals, some journalists). Saying you like an AI work means risking social contempt. Most people prefer to reject publicly out of conformity, even when their real preferences differ.

When rejecting AI is legitimate (because sometimes it is)

To be honest all the way through, we have to name the cases where rejecting AI is legitimate, because it is in several specific situations. Distinguishing the real reasons from the hypocritical ones is the important intellectual exercise.

Legitimate case 1: testimony. When we look at a work, we are sometimes specifically looking for a human testimony. A camp survivor painting her memory, a journalist writing from the front line, a writer turning her grief into a novel. In these cases, the human origin is not incidental: it is the work. AI can produce a text about Auschwitz, but it testifies to nothing. Preferring human testimony here is not hypocritical, it is consistent with what we are looking for in the text.

Legitimate case 2: the human-to-human relationship. When you buy a drawing from your favourite local artist at a neighbourhood fair, you are buying a little of her, of her time, of her way of seeing. You are supporting her life. You are creating a relationship. If the work is in fact AI, that relationship is betrayed. Not the work itself: the social contract around the work. This is a legitimate case where transparency about origin matters.

Legitimate case 3: competitions. If a competition rewards human technical skill (being able to paint, compose or write), using AI is cheating in the same way as using a calculator in a mental arithmetic contest. The rules may permit or prohibit AI, but breaking them can legitimately be penalised.

Legitimate case 4: the creative economy. As mentioned above, the economic problem (unauthorised training, unfair competition, loss of income for artists) is a real problem that deserves regulation. But it is a problem distinct from aesthetic judgement. As we explored in our article on the disruption of graphic designers, these issues require political responses, not aesthetic outrage.

Legitimate case 5: propaganda and manipulation. A fake political video generated by AI poses a real problem of manipulation, as we documented in our article on deepfakes. Rejecting AI in this context is not a bias: it is a legitimate defence of truthfulness.

There we have five cases where rejection is coherent and legitimate. But they represent a fraction of the situations in which AI is currently rejected. In the majority of cases, the rejection has nothing to do with these five cases. It is simply the Pavlovian reflex of people who cannot bear the idea that a machine might have produced something they found beautiful 10 seconds earlier.

What this hypocrisy tells us about ourselves

Beyond the debate about AI, this phenomenon reveals something important about how human perception works. Three lessons to draw from it.

Lesson 1: our aesthetic judgement is not as pure as we think. We like to think of ourselves as objective judges of beauty, who experience the work directly. The reality is that we judge a great deal on the basis of context, the label, the prestige of the author, the story told around the work. If Picasso had been a complete unknown, his works would probably be less valued than they are. This is documented by several experiments in art history.

Lesson 2: we cannot stand having our biases demonstrated to us. When you present someone with the studies cited in this article, the most frequent reaction is not "interesting, I'll revise my judgement". It is "yes, but those studies are biased" or "yes, but I can tell the difference". This epistemic resistance (epistemic = relating to knowledge) is itself a documented bias. We would rather defend our identity as "a person who judges well" than accept that we judge like everyone else.

Lesson 3: intellectual consistency is rare. If we were truly consistent in our rejection of AI, we would also reject photography, Photoshop, Auto-Tune, sampling, conceptual art, sewing machines, and many other things we accepted without batting an eyelid. The fact that we reject AI specifically, and not the rest, shows that it is not a principle we are defending, but a contextual emotion we are rationalising.

What we can do individually

A few concrete ways to loosen the grip of this mechanism.

Practise blind judgement. Before you know who made what, first ask yourself what you feel in front of the work. Note it mentally. Then read the attribution. Observe whether your judgement shifts, and by how much. This simple exercise lets you measure your own bias.

Separate aesthetic judgement from ethical judgement. An AI work can be beautiful and have been produced under questionable economic conditions. The two are not mutually exclusive. Recognising beauty is not endorsing the mechanics of production. This distinction is what separates a nuanced mind from a binary one.

Own your hypocrisy when you have some. If you use ChatGPT for your emails, don't pour scorn on "AI art". Or else explain why what you do is different. Often there is no good explanation. Recognising your contradictions is the beginning of intellectual coherence.

Criticise works for what they are. If an AI work is empty, mediocre, of no interest, say so. But say so for reasons internal to the work, not because it is AI. And conversely, if an AI work is technically accomplished, aesthetically interesting, emotionally moving, have the honesty to acknowledge it. Aesthetic judgement deserves that rigour.

Accept that we won't always know any more. In 2026, in 90% of cases where you see an image, hear music or read a text on the internet, you won't really know whether it is human or AI. This opacity is likely to increase. Learning to judge without that information will become increasingly necessary. And probably liberating.

The most uncomfortable mirror

Ultimately, what AI holds up to humanity in 2026 is not so much a technological challenge as a psychological mirror. It shows us that our judgement is less free than we think. That our tastes can be swayed by labels. That our rejection of the new looks strangely like that of previous generations faced with innovations we now accept as obvious.

This mirror is uncomfortable. But it has one virtue: it forces us to be a little more honest about our own psychological mechanisms. As we explored in our article on what makes a work human, some of our preferences for human work are legitimate and profound. Others are just social conditioning. The difference between the two is precious.

The collective hypocrisy around AI will eventually fade. Not because we will have been right or wrong, but because the generation growing up with AI today will not have the same reflexes as ours. The children of 2030 will treat AI as we treat photography: one tool among others, capable of producing good and bad alike, judged on the result.

In the meantime, it is simply useful to know that when you find a work worthless because it is AI, and you would have found it brilliant if it were human, it is not the work that is the problem. It is your relationship with the world. That is uncomfortable to admit. But it is probably the beginning of something fairer, both for you, for human artists, and for the technology that is transforming our era, whether we like it or not.

To go further, you can read our article on art in the age of AI, which explores a complementary dimension, and our analysis of the counter-movement towards the real, which shows how AI could paradoxically reinforce what is truly human. The question is not whether you like AI. It is whether your judgements about AI stand up to honest scrutiny. Often, they do not. And that is OK. It just means there is work to be done on yourself.