Artnet presents a series of stories spotlighting the women shaping today’s art industry. Paired with insights from Hardwiring Change, our second annual survey on gender equity produced in partnership with the Association of Women in the Arts, the series offers a closer look at the barriers women continue to face—and the changes they are driving across the sector. Download the full report here.
Where were you when you realized the art world was all-in on A.I.?
I was sitting on stage at an industry conference last September, when I polled the audience about their use of the technology. A sea of hands went up. I had been writing about A.I. for the better part of a decade, when it still felt like speculative fiction, but the mainstreaming of user-friendly large language models had clearly changed the picture.
The technology is revolutionary, and concerns about how it could upend work as we know it are no longer hypothetical. A growing body of research suggests that A.I.-driven automation is hurting women, who are more likely to hold jobs that will be disrupted by A.I. and, at the same time, less likely to be early adopters of the technology. The response has included pundits warning that women who fail to embrace A.I. risk getting left behind, and outright proselytizing about A.I. as a liberating tool for overburdened women. Skeptics have christened this the “girlbossification” of A.I.—the latest attempt to sell technological salvation to women.
Courtesy of AWITA.
Either way, these trends should trouble an art world already beset by gendered structural inequities, from pay gaps to the chronic undervaluation of women artists. New findings from Artnet and the Association for Women in the Arts’ second Hardwiring Change report underscore the stakes. We heard from more than 2,000 arts workers, predominantly women, and found that nearly 58 percent had experienced structural barriers in their careers related to gender, race, class, or care responsibilities, with gender the dominant factor. At the same time, A.I. uptake is already widespread: 62 percent reported using A.I. tools at work, with ChatGPT being the most common entry point (57 percent).
Against this backdrop, I was less interested in becoming either an A.I. alarmist or evangelist than in looking at how it is already being deployed across the arts, and what that might mean for women, for better or worse.
Advertisement for Whirlpool Washing Machines, circa 1953. Technological advancements have often been marketed to women as a liberatory timesaver. Photo: Fotosearch/Getty Images.
Inputs and Outputs
Commercial firms are leading the charge when it comes to A.I. innovation in the arts, and as such it is one of the areas we can most clearly see how the tech is already interfacing with the industry’s structural barriers. The market’s A.I. gold rush is in full swing, with a blizzard of start-ups offering bespoke solutions for discovery, authentication, and valuation.
Caroline Taylor, founder of the Appraisal Bureau, which uses generative A.I. to process and surface comparable works from its archive of auction and gallery data, is cautiously optimistic that the A.I. era will bring greater transparency to the art market, which could make it easier to track gender disparities. But when it comes to the business of appraisal, she was level-headed about the limits of what technology can fix.
Caroline Taylor, founder of Appraisal Bureau.
“The models are absorbing the inputs,” Taylor said. “If there’s gender discrimination structurally in the market, which there historically has been, that’s the input. So you’re going to see that as the output, unfortunately.”
The firm’s data, which is reviewed by human appraisers, shows that average fair market values for male artists run approximately 45 percent higher than for female artists. Works by men also make up around 70 percent of appraised works.
Roughly a fifth of Hardwiring Change respondents cited “bias or exclusion in datasets” among other concerns around A.I. It’s easy to see how the technology, while enabling quicker and more accurate appraisals, could also accelerate the art market’s existing feedback loops.
The problem carries over to art discovery, too. A growing array of A.I. apps and recommendation engines are combining market data with data on institutional shows, publications, press coverage, and social media metrics to surface artists to prospective buyers and confer confidence ratings.
Courtesy of NALA.
While tastemakers and curators have long influenced visibility in the art world, when A.I. powers up those same filtering mechanisms, the risk is that there will be even less room left to nourish serendipitous encounters, collectors with eccentric preferences, or emerging artists who fall outside of established circuits.
“The art market has always been structurally biased, so if you mirror that, you scale the bias,” Penelope Sonder, COO of the discovery platform NALA, told me. Her proposed solution? “You have to design against it.”
NALA’s algorithm, which originally began as a class project of founder Ben Gulak while he was a student at the Massachusetts Institute of Technology (MIT), deliberately excludes the legacy “gatekeeping” datapoints. Instead, it uses data science and machine learning to analyze visual preferences and understand a user’s taste, in order to match them with works based on “visual and conceptual affinity,” regardless of an artist’s career stage or institutional validation. With 18,000 living artists already on platform, it’s one promising experiment in how A.I. could bring about new models for art world visibility.
Designing Against Bias
Designing against bias is a challenge museums, too, will need to confront as they explore A.I.-powered tools for engaging visitors.
The Palace of Versailles recently partnered with OpenAI and culture platform Ask Mona to allow visitors to “chat” with 20 of its garden’s statues, a robot is giving tours of Turin’s Palazzo Madama, and Gen A.I. startups aimed at guiding encounters with artworks, such as Artlas and Contxt, are lining up to partner with institutions.
R1 giving tours at Palazzo Madama, Turin in April 2026.
But these tools must be created carefully. “Plug-and-play” large language models (LLMs) like ChatGPT, which have been trained on vast tracts of internet text, are already known to reproduce bias, flatten nuance, and generate confident inaccuracies. Even using machine learning for seemingly neutral tasks like image labeling can become problematic. In their project ImageNet Roulette, artists and researchers Kate Crawford and Trevor Paglen exposed harmful stereotypes encoded in the image dataset that is foundational to many deep learning models. As they argued, “the act of collecting, categorizing, and labelling images is itself political.”
Of course, this was a problem for museums long before machine learning came along. For Isabelle Mattar, author and founder of The Museum Label Project, the biggest opportunity presented by A.I. in this realm is for experimentation at the level of system design. Rather than relying on generic LLMs trained on dominant art-historical discourse, museums could build tools that deliberately surface overlooked scholarship, present multiple interpretations of a work, or contextualize how gendered and colonial biases have influenced the canon.
Beyond these training inputs, curators should also think actively about what kind of narrative voice a model adopts when speaking to visitors, Mattar said. Without direct intervention, LLMs will likely default to “the authoritative voice of a generic omniscient narrator, and, unless otherwise trained or explicitly denounced, a ventriloquized white male.”
A visitor looks at paintings in the Metropolitan Museum of Art in New York. Photo: Timothy A. Clary/AFP via Getty Images.
“I firmly believe the next challenge for museums will be to identify and personalize the voice that speaks to their visitors in galleries,” Mattar said. She added that she sees this as a crucial moment for art museums to reject the “fallacy” of an objective art history and “admit to their visitors that the decades of compounded, esteemed judgment of previous scholars does not have to inform their own appreciation—or distaste, or disinterest—for an art object.”
Museums have historically been slow to adapt to change. The Hardwiring Change data showed slightly lower A.I. uptake among workers at museums and cultural institutions (66.7 percent) compared to other sectors such as media and communications and the commercial art market, where more than three-quarters of respondents report using A.I. tools. It’s a problem leaders are well aware of. For Fine Arts Museums of San Francisco director Thomas Campbell and cultural commentator András Szántó, one of the biggest takeaways from 2025’s Museums of Tomorrow roundtable was a clarion call: If we stand idly by, “the values of the art museum will not be encoded into A.I. systems,” they wrote. The question remains whether institutions, battling against risk-off cultures and limited funding environments, will have the willingness to do this work.
A Leadership Choice
It occurs to me that most of the people who had stuck their hands up at the art business conference last September may not have been thinking about A.I. with this level of sophistication. They were probably just using ChatGPT to write their emails. Yet leaders in the arts should know that allowing this tacit embrace of the technology is not a neutral choice for women.
Exhibition view of “Kate Crawford, Trevor Paglen: Training Humans” Osservatorio Fondazione Prada. Image courtesy of Fondazione Prada. Photo: Marco Cappelletti © Trevor Paglen.
On International Women’s Day, in March, I was alarmed to read a new research brief from the International Labour Organization warning how female-dominated occupations are almost twice as likely to be exposed to generative A.I. This is particularly relevant in the art world, where women are strongly represented across many roles. The report cautioned that women’s higher exposure reflects “entrenched occupational segregation and the systemic barriers that sustain it” and that “the policy choices made now will determine whether Gen A.I. drives greater equality or entrenches disparities in the world of work.”
Just 13 percent of Hardwiring Change survey-takers listed job insecurity as an A.I. concern, making it the second-lowest ranked concern for the art world. In fact, many respondents saw A.I. as bringing the potential to relieve them of structural burdens such as “invisible labor,” or unpaid office housework, that tend to disproportionately fall on women. When asked where A.I. could be most useful, 57 percent pointed to administrative tasks, suggesting a desire to reclaim time lost to this kind of work. Nearly 70 percent said they spend more than a quarter of their week on administrative or logistical work, with 41 percent describing taking on this work outside of their formal role.
A woman working remotely on a laptop Photo: Steve Christo via Getty Images.
In one view, A.I. presents a real opportunity to redistribute that invisible labor, which could meaningfully accelerate women’s advancement. But realizing that unburdening depends on arts workplaces making access equitable, and on rewriting jobs for the A.I. era—rather than simply eliminating them. This survey suggests they are not there yet. Almost half of respondents (47 percent) said their organization lacks adequate A.I. tools or training, while a further 30 percent were unsure whether any provision existed at all. I reached out to a number of large employers in the art world to ask about their policies on the matter, and, perhaps tellingly, did not receive a clear answer. If indeed employees are being left in practice to figure out these systems on their own, then it follows that those who already have more time, money, and digital literacy, will likely be advantaged over those knees-deep in invisible labor.
All told, 43 percent of participants in the survey are unsure whether failing to adopt A.I. will worsen existing inequalities. But one thing seems for sure; it won’t automatically improve them. And as the sector engages with the technology, those in decision-making positions have the opportunity to influence the design for a more equitable future.
