Can AI Help You Buy Art?

Guaranteed to elicit passionate—and often contradictory—reactions, artificial intelligence (AI) is a hot-button issue for every industry, and the art world is no exception.

While the immediate response for some might be to dismiss the conversation altogether, AI is already being widely used and is likely here to stay. Within a creative field like the art industry, debates abound about the potential of the technology to shape and influence production and expression. Still, there are several ways that AI can streamline the art buying process.

Buyers, dealers, and advisors alike increasingly rely on the tool’s ability to gather a significant amount of information at a fast rate. But there are also limitations to using AI: It lacks the human element that can instill confidence in buyers and help develop taste. And with the vast amount of data AI can search, inaccuracies are common, not to mention the carbon emissions of AI tools, which are increasingly coming under scrutiny.

Whether you are for it, against it, or agnostic to the tool, here are some pros, cons, and caveats of using AI as an art buyer.

AI is undeniably proficient in searching for information and does so with incredible speed. A request that might take several days to find an answer can be resolved in just seconds with AI. Art buyers are already using the technology to ask questions about topics including art historical information, auction results, upcoming exhibitions an artist might have, and noteworthy accolades like awards and press.

Too much data, however, can be overwhelming to sift through, and AI can occasionally provide misleading information. Karen Boyer, founder of the advisory firm Elements in Play, recalled a recent scenario: “A client bought a painting for $200,000 that had been exhibited in the Venice Biennale, arguably the most important exhibition in the world,” she explained. “He later asked AI about pricing, and it came back with a range of $7,000 to $88,000. On the surface, that looked like a huge discrepancy, but those results were for small works on paper or uncharacteristic minor paintings, nothing comparable in quality or importance to the Biennale piece.” AI can struggle to make such distinctions.

Moreover, the technology only collates publicly available data, such as auction results. While it might be able to search in deeper corners of the web to scrape more information than a human can, it can’t gather unpublished information. As a tool for collectors to research pricing, AI is more useful to determine a general range for works by an artist in question. If a buyer is looking for art under $10,000, for example, AI can help weed out artists whose markets are clearly not a fit.

Another caveat is that the information AI provides can be outdated and incorrect. “AI aggregates information from the broad internet, and as we know in this information age, there is also a lot of misinformation,” said Mindy Taylor Ross, owner of the consulting agency Art Strategies LLC. “When asked to write a biography for Robert Rauschenberg, my mentor, AI wrote that he had served in the Marines. He served in the Navy. This is a small example, but it is about a very famous artist.”

While fact-checking and editing require time and research, for some people, this process is less arduous than facing the blank page problem of knowing where to start.

For The Melrose Gallery, which utilizes AI for various tasks, such as timing a press release, the tool can be useful for improving existing practices rather than replacing them. “The main caution is not to depend on it too much,” said Tracey Strydom, who works at The Melrose Gallery. “AI works best when it supports human insight. A good practice is to give very specific and personal information, so the results do not feel generic, and then always review and shape the final outcome with care.”

For collectors, AI can guide initial research on artists or artworks that they may want to collect.

I tried my own prompt using the basic, free version of OpenAI’s ChatGPT: “I’m interested in buying contemporary art. I really like Caravaggio,” I wrote. “Can you recommend similar artists working today?”

In the results, I could immediately see limitations with the tool. First and foremost, ChatGPT made assumptions about why I liked Caravaggio instead of asking me, which a good salesperson might do.

Some of the results ChatGPT offered were on track with what I do like about Caravaggio’s work: moody, figurative paintings with strong contrasts of light and dark (chiaroscuro), moments of physical and emotional intensity, and scenes related to debauchery, all without heavy impasto. Included in the contemporary artist suggestions it offered were Nicola Samori and Roberto Ferri, but artists I felt would be obvious, such as Jesse Mockrin, were missing. With other suggestions, the tool was more off the mark: Instead, it focused on terms associated with Caravaggio, like emotion and intensity, offering Sam Messer and Gottfried Helnwein.

With any tool, more specificity can yield better results, but very often, newcomers are at the early stages of honing their taste and don’t know where to start. It can be difficult to put into words what it is about an artist or artwork that catches our attention, in particular with technical terms like chiaroscuro.

With that in mind, it’s advisable to give as much precise information as possible with your inquiry, including your budget, any preferences like material or size of a work, and any particular factors that matter to you, like an artist’s background. When forming an inquiry, ask yourself what it is about a particular artist or artwork that you like. Is it the medium? The colors? Something about the way it makes you feel? Telling AI what you don’t like is also helpful to better tailor your results.

If you can’t put words to what you’re drawn to, AI can help with that as well. Using Caravaggio again, I asked ChatGPT to help me pin down what it is about his work that I liked, and the tool gave me a very approachable analysis of the artist’s work and the characteristics that I might be drawn to. Using these, it gave me a set of questions to better home in on the qualities of his work that I like.

After further prompting ChatGPT with additional requirements like my budget, I had a list of artists that matched my desired style and price range. Surprisingly, the tool was also able to give me contacts at galleries to reach out to, as well as templates for my outreach that it felt would garner a positive response. One approach even suggested that I should tell the gallery that I would be “happy to discuss potential museum loans in the future,” a common tactic used by collectors to gain access to a coveted work of art.

Unfortunately, the contacts ChatGPT offered were not always accurate. Also, just because I have the name of a salesperson and an email template doesn’t mean I will be offered work to buy, or that the artists even have works available to purchase.

When buying art, it’s important to consider additional factors like the provenance and condition of an artwork, which AI can’t do for you. During a transaction, there will also be points where a human touch is necessary to finesse negotiations. Buyers may need to talk through their decisions to feel confident in a purchase. While a buyer can certainly have such conversations with AI, the emotional element of collecting is better served by a salesperson or advisor. There are also post-sale factors like coordinating shipping and installation that the tool might be able to streamline, but certainly can’t replace entirely.