When Sculptures Speak: How AI Gave a Voice to the Gardens of Versailles

A partnership between Ask Mona, the Palace of Versailles, and OpenAI has turned a walk through one of the world’s most visited gardens into a real-time conversation — letting visitors speak directly with the statues that line its paths.

Imagine walking through the gardens of the Palace of Versailles on a sunny afternoon, stopping in front of a marble figure you don’t recognise, and simply asking who it is. Now imagine the statue answering — in its own voice, with its own personality, ready to discuss its origins, its creator, its place in the mythology of the palace.

That is exactly what Ask Mona co-founder and CEO Marion Carré presented at the Museum Digital Summit 2026: an experiment in conversational AI interpretation that has already generated more than 100,000 visitor messages in its first six months.

From Chatbot to Conversation

Ask Mona has been building AI experiences for cultural organisations for nearly a decade, working with more than 200 institutions around the world. The company’s founding premise — that artificial intelligence can help everyone learn at their own pace in museums — had previously produced text-based tools for cultural interpretation. The Versailles project pushed that vision significantly further.

The experience, developed in partnership with the Palace of Versailles team and OpenAI, runs across 20 sculptures in the gardens. Visitors scan a QR code beside a statue and are immediately connected to a real-time vocal conversation with that artwork. There is no app to download, no button to press and no need to look at a screen. Visitors simply talk.

“You can just walk and talk, or just look and talk,” Carré explained. The frictionless design was deliberate — the team wanted the experience to feel genuinely inclusive, accessible to anyone regardless of their familiarity with digital tools or their comfort in cultural spaces.

Building Personalities, Not Just Information

One of the most distinctive features of the project is that each sculpture has its own voice and its own character. Some are measured and serious. Others are playful. When one visitor in Carré’s demonstration asked “Hey there, who are you?”, the response came in the voice of Latona — mother of Apollo and Diana — announcing herself in triumph over the waters that had once defied her.

Creating those personalities was one of the most demanding parts of the project. The Ask Mona team worked closely with the curatorial department at Versailles to build a detailed knowledge base for each sculpture, drawing exclusively on content created by the museum. The AI was designed to be what Carré described as “trustworthy” — meaning it would only draw on that curated material to answer questions, keeping the experience anchored in scholarly accuracy even while adopting a conversational, sometimes humorous tone.

Each persona had its own editorial brief: origin, character traits, historical context, and the myths and stories specific to that sculpture. Ask Mona acted as a bridge between the Versailles team, who held the curatorial expertise, and OpenAI, who provided the real-time conversational voice technology. The collaboration involved curators, the communications department, the audio team, and digital specialists working together throughout development.

What Visitors Actually Asked

The project generated an unusually detailed picture of visitor curiosity. Because every question was logged, the team could see for the first time what people actually wanted to know when they stopped in front of a sculpture — something that is almost impossible to measure in a conventional museum environment.

Around 85% of questions were directly relevant to the sculptures themselves: their origins, their creators, the materials used, and their historical and cultural context. A further portion asked broader questions about the gardens, the palace, and the history of the French monarchy — suggesting that the experience was successfully extending visitor curiosity beyond the immediate artwork. Fewer than 1% of interactions involved attempts to test the guardrails or ask off-topic questions, a figure Carré cited as evidence that thoughtful design can pre-empt most misuse without heavy-handed restrictions.

The design of the experience accounted for the full range of conversational possibilities. If a visitor asked something tangential — say, whether a fountain statue found it cold to stand in water all day — the AI would engage with the playfulness before steering the conversation back to the artwork. If a question fell entirely outside the scope of the experience, it would explain that clearly and suggest a more relevant direction.

Breaking Down the Symbolic Barrier

One of the things Carré returned to repeatedly was the way voice interaction changes the psychological relationship between visitor and artwork. In a conventional museum setting, visitors often hesitate to ask questions — worried about appearing ignorant, or not knowing who to ask. A conversational AI changes that dynamic in ways that text-based tools do not.

“When you’re talking, you have more genuine questions that maybe you might not have asked someone,” she said. “Here you can know that it’s a tool, so you can ask any question without the fear of feeling judged.” The result is that visitors are more likely to express authentic curiosity — and more likely to engage deeply with what they are looking at.

The stories that came back from the ground reinforced this. A group of grandmothers visited Versailles specifically to use the experience, moving methodically through the gardens to find and speak with every available statue. Children who used it at Versailles began asking their parents why they couldn’t talk to the artworks at other museums. A child Carré overheard in a restaurant on the way to her summer holiday was telling his parents the AI experience was the highlight of his Versailles visit.

Lessons for Scaling

The project was built quickly — conceived at the beginning of the year and launched in time for summer — and that speed was itself part of the learning. Starting with a limited number of artworks allowed the team to be agile, test assumptions, and iterate without the risks of a large-scale launch.

One area the team underestimated was physical signage. Even with strong press coverage, visitors on the ground needed multiple touchpoints to discover the experience — a single poster was rarely enough. Making sure the experience was visible at various points throughout the gardens turned out to matter as much as the digital design itself.

Language was another gap. The experience launched in three languages, but the volume of messages received from visitors who spoke none of them made clear that the demand extended well beyond those three. Expanding the language offering is now a priority, alongside adding more sculptures to the programme.

Ask Mona is clear that the goal is not to replace human guides. “The best experience possible in a museum is to have a museum guide,” Carré acknowledged, “but if we’re pragmatic, you can’t have a museum guide for each visitor.” What the Versailles project demonstrated is that conversational AI can provide something meaningfully different from existing digital interpretation: a personalised, patient, always-available interlocutor that meets each visitor at their own level of knowledge and curiosity.

For cultural institutions considering similar projects, Carré’s advice was to start small, involve all the relevant teams from the outset, and resist the temptation to wait until everything is perfect before launching. The statues of Versailles have been standing in silence for centuries. It turns out they had quite a lot to say.

Marion Carré is co-founder and CEO of Ask Mona. She presented this work at the Museum Digital Summit 2026.