Monet painting posted as AI art exposes how we've learned to see machines

Someone posted an authentic Claude Monet painting to Twitter this week with a caption claiming it was AI-generated. Replies flooded in with confident identifications of "obvious" algorithmic tells — blurry edges, "unnatural" color blending, compositional quirks attributed to diffusion noise. Commenters pointed to the loose brushwork and atmospheric haze that defined Monet's technique as evidence of machine-learning failure. Several used phrases like "you can always tell" and "the hands give it away," despite the painting containing no hands. Others cited "weird lighting" and "inconsistent perspective" — hallmarks of Impressionism that predate Stable Diffusion by more than a century.

The stunt lands as open-weight image models have flooded social feeds with synthetic output. FLUX, SDXL, and Pony-based fine-tunes now produce photorealistic and painterly results indistinguishable from human work in many cases, training users to scrutinize every image for artifacts. That scrutiny has created a new genre of false positives — real paintings, photographs, and illustrations dismissed as "obviously AI" based on stylistic choices that long predate neural networks. Monet completed hundreds of works between the 1860s and his death in 1926; his Water Lilies series alone spans more than 250 canvases, many featuring the soft focus and color diffusion now associated with latent-space sampling. The original poster has not revealed which specific Monet was used, but the replies suggest it was one of the hazier late-career pieces — exactly the kind of atmospheric work that would trip modern AI detectors.