¹ú²ú¸£Àû Õæʵ

Stephen Smith, 56, a UK-based art collector, has bought prints through MyArtBroker, a sales and management platform for secondary-market prints and editions, since 2024, when he purchased Echo by British abstract artist Bridget Riley for £7,500 ($10,300). He uses the platform’s instant valuation tool, finding it helpful because he’s overpaid when buying from galleries before. “It gives quite a good, accurate impression of the price for a specific print at that particular time,” he says.

It’s notoriously difficult to know how much art should cost, because many galleries do not disclose prices and sales by galleries and dealers are rarely made public. The only publicly available sales information comes from art resold in the auction market. This can be frustrating for buyers, who either don’t feel confident enough to make purchases or, like Smith, say they’ve overpaid for art. More than half of collectors surveyed for the said the market’s lack of transparent pricing was a major barrier.

In response, more companies have developed artificial-intelligence-enabled valuation tools for collectors to obtain price estimates for individual artworks. So far, they seem to be increasing the confidence of new buyers, particularly at lower price points, and enabling existing collectors to track the value of their collections, according to collectors and art market specialists. Some companies are also using these valuation models to offer collectors loans backed by their art or insurance for their art collections.

Artscapy, a collection management, sales and art finance platform, offers free price estimates for collectors, together with paid appraisal services and access to art advisers. “We started using AI for valuation and appraisals with the very early models of ChatGPT, so we’ve been training the model for three and a half years,” says Artscapy Chief Executive Officer Alessandro De Stasio.

At Artscapy, the ChatGPT model is only trained on specific data sources — art sales at auction, sales on Artscapy’s platform, verifiable external private sales and data from Artscapy collectors — as well as on the company’s own algorithm. The latter also enables members to track the value of their collections and supports the company’s art loans and art insurance offerings.

Artscapy’s free AI estimates are based on this model. When a collector uploads information and photos for an artwork, the company’s £59.99 appraisal service produces a full report with the estimated value of the art, plus its likely liquidity and volatility.

Meanwhile, MyArtBroker uses predictive machine learning within its own valuation tool rather than relying on generative models to provide valuations. Managing director Charlotte Stewart says that if users want an AI valuation, they can just ask ChatGPT how much something’s worth and it will scrape all the information available in its data set. “The problem is that there’s more dirty data in this market than ever before,” she says. For example, thousands of different iterations of Andy Warhol’s Marilyn prints exist, but those could be originals, authorized reproductions or even posters.

MyArtBroker’s algorithm incorporates prints and editions sales from 400 auction houses, plus private sales and current collector demand on its platform. It also weighs 40 other factors that can affect a work’s value, including color, signatures and paper type using supervised learning, where human experts audit incoming data. “In every part of the process, you’ve got a human being in the loop,” Stewart says.

If someone using the free valuation tool selects Andy Warhol, for example, and one of the artist’s Marilyn prints, they will receive a price range for the fair-market value. Collectors whowant a precise valuation can request an expert one from a company specialist, who’ll incorporate the algorithm but consider other variables such as condition, rarity and provenance that are still better assessed by humans.

One significant headache for the field is how much of the art market’s data is missing. More than are made by galleries and dealers behind closed doors. It’s why many tool providers rely on auction sales data and private sales on their own platforms. “A big question for the market is how to access private art transactions at scale, and that’s very hard,” De Stasio says.

Artnet, a platform that owns the art market’s most influential auction price database, is also considering this question. In 2025 investment company Beowolff Capital and bought a controlling stake in Artsy, the leading online art marketplace.

Beowolff CEO Andrew Wolff wants to develop AI-enabled data and analytics that will combine Artnet’s auction market data with information on Artsy’s primary-market sales and user behavior. While the exact details are to be determined, Wolff sees many use cases for valuation tools built upon Artnet’s and Artsy’s data, including for people buying and selling art through both marketplaces and for support services, such as art lending or insurance.

AI’s image recognition and matching capabilities, as well as its ability to search vast amounts of data from museum, gallery shows and influential Instagram accounts, can rapidly enhance the amount of credible information about artworks within valuation models, particularly in the primary market. And Wolff says Artnet and Artsy’s new models will incorporate these inputs, including exhibition histories, scholarly citations, news articles, and social media and search trends. “There’s lots of structured and unstructured art market data that can inform these models, and we are starting to develop these heuristics,” he says.

But these valuation tools have clear limitations. No one interviewed for this article thought they could replace the need for art advisers, appraisers and other specialists to help collectors navigate an opaque market.

“AI can interpret visible data, read auction records and look at an artwork’s basic attributes, like the year, size and medium,” says art adviser Aileen Agopian. She thinks that’s a valuable starting point for collectors.

“You need to understand how this particular work matters in an artist’s broader body of work, its exhibition history, provenance and condition,” she says, explaining that even two works from the same artist, the same year, with the same medium and scale, can sell for very different prices at auction. Collectors can learn how these nuances affect value from advisers and appraisers and by talking to gallery and auction house staff, she says.

Much can be lost when art is reduced to a number, according to Mike Profit, CEO of ArtLogic, a business management software provider to galleries, artists and collectors. He says newer AI tools clearly benefit the industry if they increase buyer confidence, but “they negate the story behind the art and the gallery’s relationships with collectors and expertise, which in the primary market are fundamental drivers of the business.” He says that can skew a collector’s perception even before they see an artwork.

The art market’s illiquidity remainsa hurdle. Collectors with instant valuations for artworks cannot necessarily buy or sell, at that moment, or at that price, even for prints from large editions. “I trade currencies, where there’s always a [bid/offer] price,” says Smith, the collector. “These are theoretical values.” He thinks AI helps new collectors build the confidence to buy, particularly at lower prices, but says he always calls a MyArtBroker specialist before buying.

So while these tools offer a way to navigate an opaque and complex market, they’re generally the first step in the buying process, particularly for unique and high-value work. The art market is still one that’s driven by human emotions, relationships, trust and expertise. “Anyone who works in the art industry should be a humanist at their core,” De Stasio says.

Photograph: ‘Marilyn’ (1967) screenprint on wallpaper is displayed as part of the exhibition preview of “Andy Warhol. Advertising Of The Form” at the Fabbrica del Vapore on October 21, 2022 in Milan, Italy. Photographer: Pier Marco Tacca/Getty Images Europe

Topics InsurTech Data Driven Artificial Intelligence