The Human in the Room

by | Nov 16, 2026 | Art Market

It’s Coming for You! could well be the headline of many technological articles we’ve read in the past year or so. The “it” is Artificial Intelligence. Computer-directed robots have long been major parts of the factory assembly line, but the work of white-collar workers and researchers is increasingly being taken over by AI. Nowadays, even professional analysts and technical writers are uneasily looking over their shoulders.

Art appraisers are not immune to this worry. At the recent annual conference of the Appraisers Association of America, more than one panel was devoted to the use and abuse of AI in our field. The most dystopian nightmare is that a program would be invented that would enable a anyone to place an accurate value on a work of art.

In some ways, the idea doesn’t seem that far-fetched. An AI program such as ChatGPT could be fed information about buyer preferences that appraisers have long known. For example, paintings with nudes are harder to sell than paintings of clothed people. A painting of a cute puppy will attract more buyers than a painting of raccoon roadkill. A portrait of a famous person will fetch a higher price than a portrait of a provincial banker. A painting with a subject that is atypical for a particular artist will not find as many buyers as one of his or her “trademark” pieces does. And so on.

In theory, you could photograph a painting and input the image with the painting’s dimensions, its medium, its date, and so forth, into the program. AI would search its database of auction prices, adjust for size, desirability, and other variables with the work it was being shown, and then apply such factors to assign a fair market value for the piece.

Something like this has been contemplated for at least a decade. Lawsuits about proprietary intellectual property have killed such efforts so far, but we’re still in new legal territory. So is it just a matter of time before human appraisers join the ranks of the unemployed?

As Mark Twain once cabled to the New York newspapers, responding to rumors that he had died on a visit to London, “Reports of my death are greatly exaggerated.” Appraising art is an art in itself. The appraiser puts a price on something with no inherent value. Quality, that elusive trait, plays an important role, in addition to matters of condition, the current popularity of a subject or style, the venue in which a work is to be sold, the prices brought in private or gallery sales, a knowledge of planned events (such as major museum retrospectives) which will often affect the market, and many other factors. I have to weigh all of these things when setting a value.

Even such seemingly cut-and-dried factors such as auction records have to be interpreted. The auction databases don’t rule on the authenticity of the works sold; they simply list the results reported by auction houses. Many times, reviewing auction results, I have seen a painting listed as being by a famous artist sell for a result greatly below what I would have imagined a work by that artist to bring. Sometimes I can tell from the image that the work is an obvious fake and that the auction house was simply taking the signature at face value. Such “auction results” need to be eliminated from the pool when establishing comparables for the work being appraised.

On the other hand, sometimes a high price at auction is an obvious outlier. It may be that the work was from the estate of a famous person, and such provenance can make the work far more desirable to collectors than the average work by that artist generally brings. Context is everything.

In addition, when I’m doing a retail replacement value appraisal, auction records aren’t enough. What have similar works sold for privately? That information is not in the databases, and I don’t see art dealers, a notoriously secretive lot, willingly posting that kind of thing online. In past appraisals, when there was a wide discrepancy between auction results and private sales, I have had to call upon friendships with dealers, cajole curators, and resort to (almost) outright blackmail to get the information I needed to arrive at an accurate valuation.

Appraisers can use AI for such tasks as crunching numbers, which can assist an appraiser in formulating a long-range market analysis, but the results still need to be evaluated by the appraiser before their inclusion in an appraisal document. As was stressed to us at the annual conference, that AI program isn’t signing the appraisal, you are. An appraisal is a legal document, and if it is challenged in court, it’s not going to be the AI program that is on the stand.

In 1994 two Russian conceptual artists named Vitaly Komar and Alexander Melamid commissioned a research firm to conduct a poll of a thousand people about what they liked in paintings. Random people were asked three dozen questions, such as what’s your favorite color, what’s your least favorite color, what size do you prefer, would you rather have an indoor or an outdoor scene, do you like depictions of famous people, do you like figurative or abstract paintings, and so on. Using the answers, Komar & Melamid (their working name) came up with the following painting that satisfied the most desires of American respondents. (They would go on to give the same survey for people in other countries.)

Landscape painting of a lake with a wooded cliff on the left and a tall tree on the right. A man in a long coat stands on the shore, a woman in white and a child in red walk nearby, and two deer wade in the water at the right edge. Blue mountains and a cloudy sky fill the background.

Komar & Melamid, America’s Most Wanted, 1995

Photo by D. James Dee, courtesy the artists and Ronald Feldman Gallery, downloaded from www.artsy.net

The painting is the size of an armchair, is primarily blue and green, has an historical figure (George Washington), has people engaged in pleasant activities, has wildlife, and so on. Alexander Melamid later asserted, “We believe in numbers, and numbers never lie. Numbers are innocent. It’s absolutely true data. It doesn’t say anything about personalities, but it says something more about ideals and about how this world functions. That’s really the truth, as much as we can get to the truth. Truth is a number.”

Melamid was speaking of art with typically wry Russian humor, but in the case of appraisals, truth really is a number. It’s a number derived from the analysis of a number of objective and subjective criteria, set out by the appraiser and explained in the document. As long as people need accurate valuations for legal purposes, I think I’ll be employed.