This article is part of the Fine Arts & Exhibits special section on how the world of art and museums is pushing boundaries.
Museums exist to preserve treasures of history and creativity and to make them accessible to all. Until the pandemic, that was mostly a matter of getting people to come through the doors. But when Covid confined so many people to their homes, getting the art to wherever they were became a burning imperative.
Now, generative artificial intelligence is driving a revolution on that front. Some museums are using it to turn peoples’ phones into personal curators.
This November, the Cleveland Museum of Art is set to debut the new version of its ArtLens app, which already allows visitors to connect to the A.I. displays in the introductory ArtLens gallery. By holding up their phones to any collection item in the museum and tapping the scan button in the app, they will have access to all the information the museum has on it.
A.I. is fast becoming what the museum’s director, Bill Griswold, described in a recent video interview as “an essential tool” for museums. It is used to tag, describe and sort objects in digitized collections, to make search and discovery gamelike and to facilitate navigation. Crucially, it is increasing institutions’ capability to personalize visitors’ experience.
The question is whether this risks museums going down an algorithmic route and shrinking people’s tastes much the same way that streaming services have. Here’s what museums and archive professionals had to say.
For the Benefit of All
The Cleveland Museum of Art exists, Griswold says, to “create transformative experiences through art, for the benefit of all the people forever.” Its current plan is to reach 25 million yearly online visitors by 2027. Central to that goal, and the innovation on display in ArtLens, is the digitization of its collection, which is nearly complete.
Additionally, the museum has also generated visual descriptions for all 68,000 objects in its collection — the first 1,600 were written by humans and the rest were A.I. generated. Vision-impaired visitors can use their phones’ accessibility features to listen to these descriptions.
When perusing the collection online — or when scanning any collection item in the museum in person — visitors can access detailed information on each object’s acquisition, conservation and provenance, along with that visual description and in some instances a 3-D model.
At the Smithsonian’s National Museum of Asian Art, the forthcoming “James McNeill Whistler” exhibition acutely demonstrates the potential for A.I. to amplify the power of digital renderings.
After debuting in London earlier this year, this retrospective opens in Washington in February with a new addition: an A.I.-powered virtual reality recreation of Whistler’s fabled Peacock Room. With its remarkable collection of blue-and-white Kangxi porcelain, gilded shelving and hand-painted murals, the room is “a fulcrum” of the museum, Chase F. Robinson, its director, said during a video call. But obviously, the room can’t travel, and only a handful of people can be in it at once.
Now, by exploring the V.R. version with a headset, people will be able to virtually handle the room’s treasures and access different layers of information about them. As visitors pick objects up to examine them, generative A.I. will fill in the visual details, shadows and changes in perspective as the items are moved.
“A few years ago, A.I. was really bad at this,” Sonia Coman, the museum’s head of brand engagement and communication, said over video. Coman was hired in part because of her background in digital innovation.
“This was exactly the stuff that made people very skeptical about it. You would ask it to generate a still life and it would not take gravity into consideration or the shadows were all wrong,” she added.
Rethinking Navigation
The National Museum of Asian Art is also developing an A.I.-powered navigational tool. In a forthcoming introductory gallery, visitors will encounter avatars, or an A.I. blend of guide, greeter and docent, as Robinson described them. These are not intended to replace human museum staff, but instead to “be present when we can’t be,” Robinson said.
Personalization, Griswold said, is about “equipping visitors with the tools they need to enjoy the galleries on their own.” It is also about reducing intimidation, the Cleveland Museum of Art’s chief digital information officer, Jane Alexander, said over video.
“The average person can’t name an artist,” she said, nor do they always know what to ask a docent for help with. Artificial intelligence tools, by contrast, allow them to input their preferences, explore the collection creatively and get customized tours on their phones.
“The average person looks at a painting, unbelievably, for two to nine seconds,” Alexander said. But the Cleveland museum’s internal research has found that spending five to 10 minutes in ArtLens has led visitors to spend 30 to 60 extra minutes exploring the museum beyond.
“It always brings them back to the authentic art,” Alexander said.
A Little Help
Artlas, a California-based A.I. start-up, is pushing personalization for museums even further. Since starting its first exhibition guide for Tokyo’s Mori Art Museum in 2025, the company has worked with museums in Hong Kong, Bangkok, Venice and Paris, as well as with institutions across the United States, including the Brooklyn Museum, the Museum of Modern Art and the Institute of Contemporary Art, Miami.
Artlas essentially provides visitors with a customizable A.I. companion, trained on authoritative sources, controlled by the institution and conversant in 25 languages.
Before starting, Grace Yao, the founder of Artlas, conducted extensive research, speaking with institutions across the globe. She found that museums and their visitors don’t know each other very well. Artlas uses A.I. to analyze user questions and then produces a heat map of visitors’ progress through the galleries. The museums were surprised by the results.
At Dib Bangkok, Thailand’s newest contemporary art museum, the team is reworking museum signage after learning how often people ask where the bathrooms are. Mori Art Museum learned that, over two months, it attracted visitors speaking 19 languages — they had thought it would be more like five to six.
Across all the exhibitions for which Artlas has produced guides, Yao said that the pieces that visitors ended up liking the most were never those that the museum had chosen for the exhibition’s poster.
Preliminary research at the Toledo Museum of Art in Ohio, which is developing a similar app, has been equally revealing.
“Two thirds of people told us they don’t want to be on their phones,” Despi Mayes, the museum’s director of marketing and communications, said over video. The app is named Wander for how often people used that word in the study the museum did with XG Exhibit Management, an Indiana-based consultancy firm. Visitors were found to prefer getting lost. What they wanted most, Mayes said, was a “you are here” pin.
The idea, Mayes said, is “to create opportunities for connection amongst our visitors around shared passions” and devise experiences that are “hyper relevant.” To do so, she said, the museum’s app must be capable of the tasks for which people already use their phones: saving or sharing information ahead of their arrival, ordering food before they reach the cafe or buying products to pick up in the shop, as well as navigation.
The app needs to be instantly recognizable to visitors “as helpful and delightful,” she said.
In short, it must complement a visit to the museum, not drive it.
The post Museums Turn to A.I. to Create Personalized Tour Experiences appeared first on New York Times.
