AI Chatbot Helps $100 Thrift Store Painting Sell for $254K

An 88-year-old art teacher who bought a painting at a thrift store for under $100 in 1966 has sold it at auction for £189,200 (approximately $254,000) after Google’s Gemini AI chatbot identified it as a long-lost masterpiece by Scottish Colourist F.C.B. Cadell. The discovery, reported by The New York Times, highlights both the promise and the limitations of artificial intelligence in the art world.

A Six-Decade Mystery

Helene Plotkin, an art teacher with an undergraduate degree in art, purchased the painting at a thrift store in White Plains, New York, in 1966 for less than $100. The work, depicting a woman in black reclining on a bright red chair, hung on her wall for nearly 60 years. “I never, never thought about it at all, other than I loved the painting,” Plotkin told the New York Times.

Her son, Barry Plotkin, 60, recalled that family members occasionally speculated about the painting’s origins over the years, but no one ever sought a professional appraisal. That changed in December 2025 during a visit to his mother’s home in Florida.

The AI Discovery

Curious about the painting’s value, Barry snapped a photograph and uploaded it to Google Gemini, the tech giant’s AI assistant. According to Entrepreneur.com, the chatbot identified the bold brushwork, orange accents, Art Deco aesthetics, and lilac backdrop as unmistakable characteristics of F.C.B. Cadell, a celebrated Scottish Colourist who studied in Paris and was influenced by Matisse.

“It was amazing how much information came out of that,” Barry Plotkin said.

Gemini went further, instructing Barry to check the back of the canvas, where they found an auction marking, a canvas stamp, and a processing date. The AI also recommended specific experts to contact: Nick Curnow and Alice Strang at Lyon & Turnbull auction house in Edinburgh. The entire AI interaction took approximately five minutes.

Authentication and Sale

Lyon & Turnbull specialists confirmed the attribution. As Artnet News reported, the painting — titled Interior: The Lady in Black by the auctioneers — was dated to the mid-1920s, considered Cadell’s prime creative period. After demobilization from the army, Cadell had purchased a six-floor Georgian townhouse in Edinburgh, using the first-floor studio flooded with northern light as his workspace.

“As the story unraveled, we just got more and more excited, because this is the stuff of auctioneers’ dreams,” Alice Strang told the New York Times.

The painting sold at auction on June 4, 2026, for £189,200 ($254,000) — within its presale estimate of £150,000–£200,000. Cadell’s auction record stands at £874,000 ($1.2 million) for Reflection at Sotheby’s London in 2018.

AI’s Strengths and Limitations

The case demonstrates AI’s growing capability in art identification while also revealing its boundaries. Gemini correctly identified Cadell’s signature brushwork, recognized the Scottish Colourist affiliation, and even identified the artist’s signature in the upper right corner. However, it misidentified the sitter as Bethia Hamilton Don Wauchope, when the actual model was May Easter, another of Cadell’s favorites.

“The story illustrates how A.I. can get you started, but there’s no comparison for real expertise,” Strang told Artnet News. “The painting could have been a copy, it could have been by his cousin, Florence St John Cadell, and you can only tell so much through photographs.”

Experts in AI art authentication urge caution. Carina Popovici, founder of AI Art Recognition, told Artnet News: “I would be cautious about drawing broader conclusions about the suitability of publicly available A.I. tools for art authentication. The quality of the training data is at least as important as the sophistication of the A.I. model itself.”

Broader Implications

The discovery represents a new chapter in the well-established genre of thrift store art finds — previous examples include works by Picasso, Schiele, and Henry Moore. What distinguishes this case is that AI, rather than a human expert, initiated the identification. This suggests a future where general-purpose consumer AI tools could democratize art identification, enabling ordinary people to uncover hidden treasures without access to traditional expert networks.

Yet the case also underscores that AI works best as a first-pass screening tool. Final authentication still requires physical inspection, specialized archival research, and human expertise that cannot be replicated by publicly available models.

What’s Next

Helene Plotkin said she plans to give the auction proceeds to her sons. “I don’t want any. It’s theirs,” she told the New York Times. Her one hope is that the buyer will display the painting publicly so her grandchildren might one day see it.

One mystery remains: how the painting ended up in a New York thrift store just months after being sold at Christie’s London in 1966 for £21 (roughly $600 today). Strang said she has no explanation for the journey.

As AI tools become more sophisticated, stories like this may become more common — but as this case shows, the technology is a starting point, not a replacement for the trained eye of an expert.