How AI Is Changing the Way We Buy Art
Collectors are using AI to research art, and galleries are bringing it into their daily work. Its influence on sales is harder to measure
By the time a collector contacts a gallery about a painting, an AI tool may already have researched the artist or helped make sense of auction results. That first conversation may begin with more informed questions. It may also involve correcting information the collector has been given.
Much of the discussion around AI and art concerns works made with the technology. For collectors of modern and contemporary painting and sculpture, another development deserves attention. AI is entering the research that precedes a purchase and the work galleries do to present art. It can influence how someone encounters a painting made long before these tools existed.
Research before a purchase
The Art Basel and UBS Survey of Global Collecting 2026, produced by Arts Economics, offers a measure of this change. Of the 3,100 high-net-worth collectors surveyed across ten markets, 72% undertake moderate or significant independent research before buying art. Twenty-two percent use apps or AI tools for advice, up from 4% in 2024. That category combines apps and AI, so it cannot tell us how many respondents use chatbots specifically.
These findings show how the surveyed collectors use digital resources. Whether AI leads to more sales or better acquisitions remains a separate question. To understand its influence, we need to look at the information buyers receive and what they do with it.
Writing for Artsy in 2025, journalist Annabel Keenan tested ChatGPT by asking it to recommend contemporary artists based on her interest in Caravaggio. Some suggestions suited her taste. Others missed it, and some gallery contacts were incorrect. The tool also assumed what she liked about Caravaggio before asking her. Her experiment captures one experience with a particular version of ChatGPT.
For a collector, the useful next step is to ask for sources and read them. Do they support the answer? A recommendation may introduce an artist worth exploring, but a purchase requires research into the individual work.
Inside the gallery
The Artsy AI Survey 2026, conducted in February among more than 300 gallery professionals, records some everyday uses. Fifty-seven percent use AI to draft or edit communications, 24% for research, archiving or data organization, and 19% for installation renderings or virtual exhibition design. Thirty-one percent report no use of AI in their operations. These figures describe the respondents, rather than the gallery sector as a whole.
For a small team, help organizing material or preparing a first draft could leave more time to study works and respond to collectors. How much time it saves depends on the task and the corrections it requires. A draft that needs extensive checking may offer little advantage.
Commercial results need their own measures. Counting tasks completed with AI tells a gallery little about purchases. To assess its effect on sales, the gallery would need to follow inquiries through to completed transactions and account for the time and cost involved. The evidence reviewed here does not establish an overall increase in art sales attributable to AI.
Who appears in the recommendations
AI search gives galleries another reason to pay attention to their public information. Artlogic, a software provider for art businesses, recommends maintaining complete artist and artwork pages, exhibition archives and current availability information. Its guidance also suggests tracking referrals and inquiries to understand whether AI tools are helping people discover a gallery. This is commercial advice from a provider, rather than independent evidence that these practices produce sales. [4]
Collectors need to know where a recommendation comes from and whether the information is current. Galleries have reason to check what these tools are saying about their programs. A work described as available after it has sold, or listed with incorrect dimensions, can complicate an inquiry before anyone has spoken.
There is also a question worth investigating. Do these recommendations help people discover less familiar artists, or do they favor those with the strongest digital presence? Answering it would require systematic tests using comparable queries. A search result gives us no assurance that the artists it omits are less relevant.
The works a search may miss
An AI search depends on the sources the tool can access. A painting kept in a private collection for decades may appear in its answers because it is documented in an accessible catalogue. Another may be known through private archives and conversations that fall outside a public search. Even when a tool identifies a work, it may have no information about its availability or recent documentation.
Rarely exhibited and undocumented are different conditions. A work that has seldom been shown may have an extensive published history. One that has never been published may have no public references at all. Private records could become available to an AI system if it were given authorized access, so the limits depend on the system being used.
For collectors, the question is what information is missing and who can provide it. A gallery, archive or specialist may be able to continue an inquiry beyond the public record. Limited exposure alone tells us little about a work’s quality or value.
Price and authenticity
Researchers are also studying AI’s use in valuation. A December 2025 preprint on arXiv examined models combining auction data with images. Artist identity and previous transactions carried the most predictive weight overall. Visual information made a particular contribution for works without their own prior sales history. Those findings apply to the data studied and do not establish that the models can reliably value any work or private sale.
Before relying on an automated estimate, a buyer should know which comparable sales were used, when they occurred and what information about the work was included. A precise-looking figure can still rest on incomplete records.
Authentication involves a different kind of assessment. Specialized systems for analyzing images need to be distinguished from general-purpose chatbots. Guidelines developed in 2025 by the Center for Art Law and Art Recognition call for transparent methods, scientific rigor and collaboration with experts. They place AI alongside provenance research and examination of the work’s materials. Art Recognition also supplies authentication technology, a commercial interest readers should bear in mind when assessing the guidance.
A buyer should be able to understand what was examined and how the findings relate to the other documentation. An estimate addresses price. Attribution concerns authorship, while provenance traces ownership. Evidence for one does not settle the others.
Encountering art through AI
AI can play a role well before someone considers buying. The Cleveland Museum of Art opened ArtLens Reimagined to the public on July 28, 2026. Its Talk to the Art experience lets visitors ask questions about works using reference material approved by curators. According to the museum, the AI models run locally, and these conversations draw on a defined library of sources. This offers a concrete example of AI supporting interpretation in a museum. Its effect on future purchases would require separate research.
Art made with AI has also found buyers. Christie’s Augmented Intelligence auction, which closed on March 5, 2025, totaled US$728,784. The auction house reported that 37% of registrants were new to Christie’s and 48% of bidders were millennials or Gen Z. Those figures describe this particular sale and its participants. They cannot establish demand across the market.
The museum experience concerns how visitors engage with a collection. The auction raises questions about the reception and sale of an artistic practice. Both deserve attention, with evidence suited to what each can tell us.
The conversation about a particular work
Digital research continues alongside relationships with dealers. In the Art Basel and UBS survey, 87% of respondents had bought through a dealer in some form over the previous year, including at art fairs. The figure identifies a purchasing channel, without telling us whether AI played a part in those decisions.
At Gary Nader Art Centre, a discussion of a work by Botero or Lam involves questions specific to that painting. Where does it sit within the artist’s career? What records help us understand its history? Research into Soto or Cruz-Diez likewise requires attention to the particular work. Some answers may depend on archives or documentation absent from a public search.
A collector arriving with an AI recommendation has good reason to ask the gallery to explain its sources and the choice of comparable works. The gallery should be able to support its answers with evidence about the painting or sculpture under consideration and explain where the record remains incomplete.
