On 14 May 2026, an X user called SHL0MS posted a cropped image of a painting with the platform’s “Made with AI” label attached, and asked the internet for an honest critique.

Within hours the post had 2.3 million views, hundreds of replies, and a long list of confident criticisms. The composition was attacked. The colour palette was dismissed as “obvious AI”. One commenter wrote an 850-word breakdown of its “soulless AI shortcomings”.

The painting was a genuine Claude Monet. Water Lilies, painted around 1915, currently hanging in the Neue Pinakothek in Munich. One of the most-reproduced images in Western art. Nobody noticed.

It is, accidentally, one of the funniest things I have read in months. It is also, quietly, one of the more uncomfortable things I have read in months. And I think it is worth a longer look than the screenshots got.

The setup, and why it worked

I am not interested in piling on the critics here. The internet has done that already, with vigour, and I do not think it tells us anything we did not already know. The interesting thing is the mechanism, not the people who fell for it.

People did not look at the painting and decide it was bad. They looked at the label, decided what they were going to think, and then read the painting through that filter.

That is a normal cognitive shortcut. We use it every day. We see a name on a CV and we form an opinion before we read the experience. We see a postcode on a business card and we adjust our quote. We see a logo and we relax our scepticism on a claim we would otherwise interrogate. The label is doing the looking.

Most of the time, the shortcut is fine. The label is roughly accurate. The judgement that flows from it is roughly proportional. The world keeps moving.

What the Monet experiment showed is what happens when the label is loaded with strong emotional weight. “Made with AI” is one of those labels right now. For one group it means “lazy, soulless, plagiarised, derivative”. For another group it means “exciting, fast, democratising, future”. Almost nobody encounters the phrase without some pre-loaded reaction attached.

Drop that label on a Monet, and the painting becomes whatever the label tells you to see. Which is not, in this case, what the painting actually is.

There is even scientific research showing that the mere belief that a work was generated by AI measurably reduces the aesthetic experience of viewers. The belief moves the experience before the eyes have done their work.

Why this is bigger than art

The art world will be fine. Monet will keep selling. The X thread will become a case study in some media studies seminar in three years. The interesting question is what this tells us about the rest of our work.

Because we are doing this everywhere now, not just to paintings.

We are doing it to job applications. We are doing it to proposals. We are doing it to academic papers. We are doing it to news articles, social posts, internal reports, marketing copy, customer service replies and, recently, to entire genres of content that have been quietly produced with AI help for years.

The label “AI” is now load-bearing in a way it was not eighteen months ago. People are forming an opinion of a piece of work the moment they suspect AI was involved, before they have judged whether the work is any good. That is a problem in both directions.

It is a problem if AI was involved and the work is genuinely good, because the reaction to the label is now stronger than the reaction to the work. It is also a problem if AI was not involved and the work happens to look like something AI might have produced, because the same reflex kicks in. Either way, the label is doing the looking.

The Monet test simply exposes the reflex in its purest form.

What it does not mean

I want to be careful here, because there are a few wrong lessons to take from this story.

One is “see, AI art is just as good as the masters”. No. That is not what happened. What happened is that hundreds of people, in a hurry, on a platform optimised for fast reactions, did not slow down enough to recognise a famous painting under a different label. That is a story about the readers, not the painting. It would have happened with almost any image.

Another is “we should stop labelling AI work”. No. Provenance matters, the SynthID rollout and the C2PA work that OpenAI has just joined are good things, and disclosure is a feature, not a bug. The fact that the label is currently loaded with strong reactions is a transitional problem. The label is still the right idea.

The lesson that does generalise is gentler than either of those. It is that we should be a bit more careful about letting the label do all the work.

A practical version

Most of the people reading this are not going to be in a position where they are asked to critique a Monet under a misleading label. The everyday version of this for the people I work with looks more like:

• The candidate’s CV was clearly partly drafted with AI help, and I now do not know what to do with it.
• The intern’s report reads like AI. Is it any good?
• This LinkedIn post sounds AI-generated. Should I trust the argument inside it?
• This proposal has the formatting tics I associate with AI. Did the firm actually do the work?

If you find yourself in any of those moments, the move is the same. Treat the label as one data point, not the whole verdict. The question is not “was AI involved?” because the answer to that is increasingly “yes” for almost every piece of professional work you will encounter. The question is “is the underlying work any good, and did the human apply judgement to it?”

That is the question the Monet critics did not stop to ask. That is the question we keep skipping.

It works in the other direction too. If you have used AI to help with something and the output is genuinely good, do not let the label of “AI assistance” become the headline. The headline is the work. The label is the accompaniment.

The honest summary

The Monet experiment is a perfect, accidental piece of performance art. It is also a quiet warning about where our judgement has drifted to.

We have moved on from “can AI make art?” because that question is settled. The new question is whether people can still see the art when they have already decided how they feel about it. That question applies, in some form, to almost everything any of us is doing professionally right now.

The label is not the work. The label is not the candidate. The label is not the argument. Slow down, look at the thing in front of you, and decide whether it deserves your time before the label decides for you.

If you can do that consistently, you will be one of the people who comes out of this transitional moment with their reputation for fair judgement intact. That is going to matter more than people think.

I discussed the Monet test on Chapter 14 of Prompt Fiction, our podcast on what is actually happening in AI.