Ai in galleries: When the algorithm becomes curator

This dichotomy between AI’s media image as an artist and its operational reality reveals a fundamental paradox: galleries are massively adopting these tools, yet their integration remains largely confined to logistical rather than artistic functions. Still, a few pioneers are already transforming this technology into a genuine creative medium, redrawing the boundaries between curation, creation, and the art market.

01Behind the scenes of AI: when galleries automate the invisible

In the hushed offices of Perrotin gallery in Paris, a small team now manages the equivalent of three physical spaces and a digital platform using artificial intelligence tools. "We use machine learning models to analyze attendance data from our exhibitions," explains a manager who prefers to remain anonymous. "AI helps us optimize the hanging of works based on visitor pathways, and even predict which pieces will generate the most interest during openings."

This discreet but systematic use of AI primarily involves: Inventory management: Algorithms analyze sales histories to recommend which works to highlight during different times of the year. Thaddaeus Ropac gallery uses a predictive system that increased its sales by 18% in 2023, Flow analysis: Smart cameras (without facial recognition) map visitor movements to identify congestion zones. The Centre Pompidou tested this system during its 2020 exhibition "Global(e) Resistance." et Content generation: Tools like Jasper or Copy.ai automatically produce artwork descriptions and press releases. Marian Goodman gallery saves 15 hours of work per month this way.

"AI doesn’t replace our teams, but it frees up time for what really matters: contact with artists and collectors," explains a Paris gallery director. This pragmatic approach explains why 68% of galleries surveyed by Artprice in 2023 reported using AI without making it a marketing argument.

02Generative art: when the algorithm becomes brush

Yet a minority of players are transforming AI into a genuine creative tool. Pace gallery, always at the forefront of new technologies, opened a space dedicated to generative art in its New York branch in 2022. "We consider these works a new form of artistic expression, on par with photography or video," explains Marc Glimcher, its president.

Among the represented artists: Refik Anadol: His monumental installations, like "Machine Hallucinations" shown at MoMA in 2022, transform raw data into immersive experiences. For this work, Anadol trained a model on 113 million images of New York architecture, Ian Cheng: His "Emissaries" are generative simulations that evolve in real time, creating infinite narratives. Hauser & Wirth gallery presented his work at the latest Art Basel et Sougwen Chung: Her performances blending human drawing and robotics were exhibited at the Cartier Foundation. Her "Drawing Operations" project uses AI to create collaborative works between artist and machine.

These artists share a common approach: AI is not merely a production tool but a creative partner. "I don’t ask the algorithm to create for me, but to help me explore territories I wouldn’t have imagined alone," explains Chung.

03The art market faces the authenticity challenge

The arrival of AI in the art market raises fundamental questions. In 2018, Christie’s sale of "Portrait of Edmond de Belamy," an AI-generated work, for $432,500 caused a sensation. Yet this transaction already revealed the system’s ambiguities: The work was created by the Obvious collective, which used open-source code developed by a student, The painting bore the signature "min G max D x [log(D(x))] + z [log(1-D(G(z)))]"—the mathematical formula of the generator et No mention was made of the thousands of works used to train the model.

Since then, galleries have had to adapt their practices. "We now demand complete transparency about the datasets used," explains an art law expert. "An artist using AI must be able to prove they have rights to all images used to train their model."

This requirement has led to the emergence of new platforms like: Art Recognition: An authentication tool using AI to detect counterfeits, Verisart: A blockchain solution certifying the provenance of digital works et Codex: A decentralized database recording generative works’ metadata.

"The real challenge isn’t technological but legal," specifies a specialized lawyer. "How do you protect a work that can be endlessly reproduced? How do you compensate artists whose works were used to train models?"

04Galleries face the ethical dilemma

AI’s use in art also raises complex ethical questions. In 2023, several artists sued Stability AI, Midjourney, and DeviantArt for copyright infringement. Their arguments: AI models were trained on millions of works without their creators’ consent, Artists aren’t compensated when their styles are reproduced et Some platforms allow generating works "in the style of" without compensation.

Faced with this controversy, galleries are taking divergent positions: Templon gallery has decided to represent only artists using AI transparently and ethically, Almine Rech gallery organizes debates on the subject, like "Can AI be a tool for decolonizing art?" et Chantal Crousel gallery has created a fund dedicated to researching AI’s ethical implications.

"We can’t ignore these questions," explains a Paris gallerist. "AI isn’t neutral: it reproduces the biases of its training data. If we’re not careful, we risk perpetuating the art world’s inequalities."

05AI as a democratization tool?

Yet some see AI as an opportunity to democratize art. Kamel Mennour gallery launched a project called "Art for All" in 2023: Generative works are sold at accessible prices (between €500 and €2,000), Collectors can personalize works through a simple interface et A portion of profits goes to associations supporting art education.

"AI allows creating unique works at reduced costs," explains a project manager. "It’s a way to reach new audiences who couldn’t afford traditional art."

This approach aligns with platforms like: Art Blocks: Offering generative works as NFTs, Fxhash: An open-source platform for generative art et Teia: A decentralized marketplace for digital art.

"AI’s true potential lies in its ability to create personalized artistic experiences," explains an expert. "Imagine an exhibition where each visitor sees a unique version of the work, adapted to their tastes and journey."

06The algorithm’s limits: when AI fails to create

Despite these advances, AI still shows its limitations when it comes to artistic creation. Several recurring problems have been identified: The "uncanny valley" effect: AI-generated faces often have slightly distorted proportions, creating discomfort in viewers, Anatomical errors: Hands with six fingers, misplaced eyes, disproportionate limbs et Lack of narrative coherence: Models struggle to create works with a genuine story or clear message. Style repetition: Generative works often share similar aesthetics, marked by saturated colors and symmetrical compositions.

"AI excels at reproduction but struggles to innovate," explains an art critic. "It can imitate Picasso but can’t create a new artistic movement."

These limitations explain why most galleries still use AI as an assistive tool rather than an autonomous creator. "We use it to generate ideas and sketches, but the final work always remains human," specifies an artist represented by Nathalie Obadia gallery.

07Toward a new era of curation?

AI’s most profound impact on galleries may well be in the field of curation. Several institutions are already experimenting with algorithmic curation systems: Tate Modern tested a system generating hangings based on attendance data, Centre Pompidou uses AI to analyze visitor reactions and adapt exhibitions in real time et Cartier Foundation created an exhibition where works are selected by an algorithm based on visitor preferences.

"AI allows creating more dynamic and interactive exhibitions," explains a curator. "It can also help discover emerging artists by analyzing millions of data points."

Yet this approach raises questions about the curator’s role. "An exhibition isn’t just a collection of works," specifies an independent curator. "It’s a narrative, an experience, something that goes beyond mere data aggregation."

08The future of AI in galleries: between opportunity and threat

As AI gradually establishes itself in the art world, galleries must navigate between opportunities and challenges. Several scenarios are emerging for the coming years: AI as tool: Most galleries will continue using it for logistical tasks without making it a marketing argument, AI as medium: A minority will transform AI into a genuine form of artistic expression, creating a new market et AI as threat: Some artists and galleries could be marginalized if generative models become too dominant. AI as regulator: Institutions might use AI to analyze market trends and adapt their policies.

"The real challenge isn’t technological but human," explains a gallerist. "How do we integrate these tools without losing what makes art special: emotion, singularity, imperfection?"

One thing is certain: AI has already transformed the gallery landscape, even if its impact remains largely invisible to the general public. Like photography in the 19th century or video in the 1960s, it represents both a revolution and a threat to the art world. How galleries choose to integrate it will determine the future of artistic creation.