Forum
The British Museum’s Synthetic Visitor and the Trust Museums Trade On

On 27 January, the British Museum published a set of images on Instagram and Facebook showing a young woman looking thoughtfully at objects in its galleries. The caption encouraged followers to take time for a closer look. Viewers soon noticed that the woman was not a visitor at all. The images were AI-generated, and the post tagged an account presenting an AI model and an AI marketing agency.
The post came down later the same day. According to reports, it had been online for about six hours, long enough to draw sharp criticism from archaeologists, heritage professionals and members of the public. Critics also pointed out that the synthetic visitor appeared in outfits evoking East Asian and Mexican dress while looking at objects from those parts of the world.
It was a small campaign on social media. It touched on some of the largest questions facing cultural institutions that want to use AI.
Why it landed so badly
Museums trade on trust. People visit and follow them because they believe what they see has been looked after, researched and presented honestly. An image of a visitor in a gallery is a modest claim, but it is still a claim: someone stood here and looked at this.
Replacing that someone with a generated person quietly breaks the claim. It can suggest that the institution’s own visitors were not worth photographing, and that the public might not notice. For an organization whose authority rests on authenticity, that is an expensive message to send to save a photoshoot.
The representation problem
The costumes made it worse. A generated woman dressed to match the cultures of the objects she was looking at turns identity into styling. When real visitors from those communities exist, and when museums are under scrutiny about how they present the cultures in their collections, a synthetic stand-in reads as both careless and revealing.
This is a pattern designers should recognize. Generative models make it effortless to produce people who look a certain way. That ease can hide real decisions about representation that would have been discussed, and probably rejected, if a human casting process had been involved.
What good policy would look like
Many institutions do not yet have clear rules for AI imagery in their own communications. The British Museum episode suggests what such rules should cover.
- No synthetic people presented as real visitors, staff or communities.
- Clear labelling whenever AI-generated imagery is used, in a form that survives sharing.
- Review of any generated depiction of cultural dress, practices or identities by people with relevant knowledge.
- A preference for commissioning photographers, illustrators and designers for work that represents the institution.
- An explanation, published at the time, when generated imagery is used deliberately and for a reason.
Rules like these do not ban AI. They place it where it can serve an institution’s purpose without undermining what makes the institution worth following.
A note for creative professionals
Designers and agencies working with museums, galleries and heritage organizations are likely to be asked about AI imagery more often this year. It is worth having a position ready. Explaining why a real photoshoot, a commissioned illustration or a clearly labelled synthetic image is the right choice for a given brief is now part of the service.
Awards and the same standard
The criteria on the International AI Award Rules page ask entrants to consider fairness, inclusivity and respect for human values when using AI, and to be open about how their work was made. The British Museum’s post failed on exactly those points, which is a useful reminder that ethical criteria in competitions are not abstract. They describe the judgments audiences make every day.
The bigger lesson
Generative tools will appear in cultural communication; some uses will be thoughtful and some will be careless. The difference will not be the technology. It will be whether an institution asks, before posting, what the image claims, who it represents and whether its audience would feel misled. The British Museum learned the answer in six hours. Others can learn it more cheaply.



