How AI Is Restoring Lost Masterpieces and Changing Art History

I'm Parikshet, I'm 11, and this is the coolest use of AI I've found yet: bringing back paintings that humans lost, destroyed, or never finished. Art and AI feel like opposites — one is emotion, the other is maths. What I discovered is that the maths can understand the emotion well enough to recreate it.

The Problem With Old Masterpieces

Centuries of war, fire, water damage, and clumsy earlier restoration attempts have destroyed or damaged huge portions of the world's art. The Night Watch by Rembrandt was cut down by about 60cm on each side in 1715 to fit through a door. Michelangelo's Sistine Chapel frescoes are faded. The Mona Lisa has cracks from temperature changes over 500 years.

Restoring these properly used to mean human conservators working for years with tiny brushes, guessing at original colours and forms. AI changes the equation.

The Next Rembrandt

In 2016, ING Bank and Microsoft trained an AI on all 346 surviving Rembrandt paintings. The AI learned his lighting style (always from the upper left), his favourite facial features, the way he built up paint layers. Then it generated a completely new portrait in Rembrandt's style — one that no human painted. It was 3D-printed using height data from the real brushwork texture so it even felt like a Rembrandt.

Art historians were split. Some called it a masterpiece of technology. Others said it proved nothing about Rembrandt — only about pattern recognition. I think they're both right. It's not a Rembrandt. But it's a fascinating window into his mind.

Restoring The Night Watch

In 2021, the Rijksmuseum in Amsterdam used AI to reconstruct those missing side panels cut off in 1715. They had a 17th-century copy of the original painting by Gerrit Lundens to train from. The AI upscaled and adapted that copy to match Rembrandt's original scale and style, then printed the reconstruction to hang beside the real painting. Visitors can now see the Night Watch as it was meant to look.

Revealing What's Underneath

X-ray and infrared imaging have long shown restorers what lies beneath a painting's surface. AI makes that data readable. A 2022 project at Antwerp University revealed a complete hidden portrait under a Picasso — a woman's face he had painted over. The AI separated the overlapping pigment layers to reconstruct both images independently.

Van Gogh's Patch of Grass was found to hide a peasant woman's head underneath. AI analysis of paint chemistry confirmed it was Van Gogh's own hand beneath — not a later addition by someone else.

Catching Fakes

Forgery is a billion-dollar problem in the art world. AI trained on a painter's verified works can now detect statistical anomalies in brushwork direction, pressure variation, and pigment distribution that human eyes miss. A Hahn Picasso that had been disputed for decades was analysed by Rutgers University AI in 2021 — the system concluded it was highly unlikely to be genuine based on brushstroke patterns. The debate is ongoing, but the AI gave investigators a new angle.

Why This Matters to Me

I love that AI is not just a future thing — it is digging into the past. Every hidden painting it reveals, every faded fresco it restores, every forgery it catches means more of human history is preserved. When I think about the art that might still be recoverable with the right tools, it makes me want to study both machine learning and art history. Maybe that's the career I didn't know I wanted.