Why Can't People Tell AI Art From Human Art Anymore?

In May 2026, an anonymous artist going by SHL0MS ran a small experiment on X. They posted an image, labeled it as AI-generated in the style of Monet, and asked people to explain exactly why it looked fake. Thousands gave their two cents on why it resembled AI - from critiquing the brush strokes to the colours, they had lots to say about how off and awful the alleged AI-generated painting looked. 

Two days later came the reveal: the painting was a real Monet, part of the Water Lilies series from around 1914, hanging in the Neue Pinakothek in Munich. Humbling indeed.

That episode is such an apt symbol of where we are now with AI becoming such a burgeoning form of tech in 2026. The old assumption that AI art has a "look," and that a discerning eye can spot it has stopped being true. 

Real work is getting mistaken for fake

For most of the last few months, spotting AI-generated images was just a game of sleuthing out the tell: too many fingers, garbled text, pixels glitching out, faces that gave off uncanny valley. 

As technology advanced, this heuristic aged badly. Improvements through 2025 and 2026 closed most of the old gaps – limbs and extremities are now more often than not rendered correctly, and texts are becoming more legible. The visual tells people trained themselves to spot are slowly disappearing. While discerning users are now more confident than ever in their ability to detect AI, the chances of them being able to identify it now is lower than ever. And the unfortunate thing is that CGI artists and traditional illustrators are absorbing the collateral damage. 

A recent post by Adidas took Singapore’s online community by storm when they launched a CG video for their Adidas Orchard Ignition event with an F1 live race car demo. Multiple comments accusing the brand of using AI were left under the post, despite the post actually being created with real CGI.

What does this say about consumer sentiment in 2026? Audiences are too quick to declare something is generated by AI, but can we really blame them?

Skepticism fatigue

When a technology moves this fast, staying accurately skeptical is expensive. It requires updating your mental model every few months and admitting that you may be wrong. It's easier to keep applying the same rules regardless of whether they still hold, or to swing the opposite direction and assume everything might be fake, real evidence be damned.

Fatigue also shows up as a loss of nuance. "Is this AI?" has collapsed into a yes-or-no reflex. Professional workflows now routinely blend traditional technique with AI-assisted steps – for example, generative fill for a background, an AI chat for concept exploration, and upscaling tools integrated into traditional editing software. Treating "AI" and "human-made" as two hermetically sealed categories was already a stretch by 2024. By 2026, it's close to a category error, but the binary instinct persists because it's easier to accept by the masses.

If the eye test no longer works, the burden shifts. Audiences can't be expected to out-detect the technology, so the responsibility for using it well lands on the people making the work.

For creatives: use AI with intent, and guard the craft

The question worth asking is no longer "can AI do this?" (increasingly, it can) but "should it, and what do we lose if we use AI?" 

AI is genuinely useful for the scaffolding of creative work. Used this way, it clears space for better work rather than replacing it.

However, craftsmanship still matters most where the work needs a real perspective. The creatives who will stand out are the ones who know which parts of the process are worth protecting, and where to keep AI as a tool in their workflow.

Disclosure is the new trust signal

When no one can reliably tell, saying so becomes part of the work. A simple note on how something was made, whether it's "conceptualised with AI, illustrated by hand" or "fully human-made, start to finish", does two things at once. It gives audiences the context they're otherwise left guessing at, and it protects genuine work from the reflexive "this is AI" accusation, as the Adidas team learned the hard way.

Disclosure also works in the other direction. Creatives who are open about using AI where it makes sense tend to earn more goodwill than those caught quietly hiding it. Audiences are far more forgiving of a tool than deception. Transparency costs very little, and in an environment defined by suspicion, it is one of the few things that builds credibility.

For brands, the stakes are higher

Whatever a brand puts out is a statement of what it stands for. A campaign, a caption, a tagline: each one is a small signal of how much went into it.

The temptation is obvious. AI makes generating everything efficient, but a brand that outsources all of it slowly starts to sound like everyone else, and audiences may not be able to name exactly why, yet they'll feel it. The work comes off hollow, the voice loses its specificity, and the resonance that took years to build erodes. Once people sense that no one is really behind the content, trust is hard to win back.

This is where selectivity matters most. Brands should be deliberate about where AI earns a place and where the human touch is the whole point – like the brand voice and the moments meant to make people feel something. Efficiency is a good reason to use a tool, but it's a poor reason to abandon your brand’s point of view.

Our Takeaway

The Monet incident showed that our instincts for authenticity are now unreliable, and that suspicion, as much as deception, can do equal damage. In a landscape where anything might be questioned, the safest ground is work that's made with care, shared with real transparency, and backed by real people-based insight.