The visitor has changed. Not the tools.
Why mediation must become a dialogue again.
The premise.
« All mediation, from the wall label to the audio guide, rests on the same premise: the venue speaks, the visitor listens. »
This premise has held for fifty years. Printed labels on walls, audio guides around necks, apps downloaded at the entrance: each generation renewed the medium without ever changing the rule. The venue speaks. The visitor listens.
The visitor was listening. Politely, sometimes attentively, often less than you hoped. But listening. The venue carried the single voice. Knowledge flowed from curator to public, from panel to eyes, from headset to ears.
This premise no longer holds. The visitor doesn't listen passively anymore. They query in parallel. They query another interlocutor than you: a general-purpose artificial intelligence, in their pocket, answering in your place without knowing your venue, your work, your sources.
What's happening here is more than a shift in usage. It's a shift in who's speaking.
The visitor, today.
Today's visitor is not an audience. They are an interlocutor.
Picture this. A visitor in front of a large 17th-century oil painting in your collection. She stops. She wonders who painted it, what the scene in the background represents, why the framing is what it is. The label answers part of her questions. For the rest, she takes out her phone.
She isn't asking you. She is asking ChatGPT.
And ChatGPT answers. Without knowing your recent attribution, without knowing the work was restored in 2019, without having read your scientific team's study that proposes a different reading. It answers with what it has at hand: an average of what the web knows about works that resemble this one.
The visitor leaves with a reading. Not yours. A diluted reading, plausible, without the nuance that distinguishes your narrative from the general one.
This will not stop. The visitor now arrives with their own tools. What they expect is no longer a one-way broadcast, but a dialogue, at their pace, on what they want to dig into.
If they don't find it from you, they will look elsewhere, at tools that know nothing of your venue.
What is at stake.
The debate is bigger than your venue. Bigger than ours.
Cultural institutions are among the last sources of verified knowledge. Every word on a label has been arbitrated. Every attribution has been weighed. Every date is traceable. This discipline of arbitration distinguishes knowledge from assumption.
When a visitor asks a general-purpose AI about your venue, they don't receive your knowledge. They receive an average, calculated from sources nobody verified, weighted with others whose origin is unknown.
At scale, this doesn't only dilute your narrative. It dilutes shared knowledge.
Museums, galleries, and heritage sites are not content providers among others. They are the guardians of a knowledge that takes the time to verify itself. This status carries a social value that exceeds the activity of any one venue. It is what must be preserved.
Two bad options.
Do nothing
You let general-purpose AI assistants tell the story of your venue in your place. They don't know what your scientific team wrote. They don't know the nuance of your wall labels, the work of your mediators, the choice you made on a specific word, attribution, or date. They invent. Sometimes accurately, often vaguely, sometimes wrongly. The visitor leaves with a diluted version of your narrative, unaware that it's diluted.
Do it the old way
You commission an agency. The quote arrives: tens of thousands of euros, several months of production, full dependence on the vendor for every update. One more phrase, one mistake in the audio guide text, and the project reopens. You lose your autonomy over what matters most: your voice.
Our position on AI.
You are reading us and thinking: they criticize AI, and they sell it.
Let's be clear. We are not against AI.
We are against AI without source, without voice, without a home. We are against AI that answers in the place of a museum without knowing its work. We are against AI that speaks about art by aggregating what it has scraped from the open web.
An AI grounded in your texts, your arbitrations, your sources: that is one more tool for mediation. An AI disconnected from all that: that is a risk for the narrative you carry.
The distinction is not ideological. It is practical. The same technology can serve or betray, depending on how you use it.
Gidmi is the use we believe is right.
The third path.
A platform built for you. Not to replace you.
- Create
- You produce your own mediation. No agency, no intermediary stepping between your knowledge and your visitor. Your team keeps editorial arbitration end to end.
- Companion
- The visitor enters a dialogue with your venue, not with a machine. The conversation stays within your narrative, grounded in your sources. The companion does not leave your house, even when the visitor digs into an unexpected question.
- Control
- Every word reaching the visitor stays under your control. Your texts belong to you, editable at any moment, without asking a vendor for permission. Your visual identity, your tone, your editorial truth: your arbitration alone.
- Insights
- You learn from your visitors. What holds them, what loses them, what they never ask. Not to optimize the way a marketplace optimizes. To inform the next exhibition from concrete observations.
Accessibility as continuation.
The third path doesn't just speak to visitors already familiar with your venue. It explicitly includes those that classical tools don't reach.
Visitors with visual or hearing impairments. Visitors in easy-read mode. Visitors who don't speak your language. Visitors who need another pace, another reading level, another voice.
We design Gidmi following the requirements of WCAG 2.1 AA and RGAA, with screen reader compatibility built in. Your team can write an easy-read tour whenever your audiences need it. The Musée Louis Braille illustrates this concretely: accessible mediation is not a module added at the end of a project, it's a starting condition.
Accessibility is not an added cost. It's the same mediation, made for everyone.
Our commitment.
Three concrete commitments.
- Your AI is grounded.
- It responds only from your sources. It does not hallucinate your history. It says "I don't know" rather than invent.
- Your data is hosted in Europe.
- Hosted on GDPR-compliant infrastructure. Your texts are not used to train AI models.
- Your team keeps the keys.
- You can edit, remove, replace any tour at any time, without asking us. No lock-in, no dependence.
These three commitments are our definition of editorial autonomy.
Toward what world.
In time, every cultural venue speaks in its own voice. To each of its visitors. At scale.
The curious visitor enters a conversation with the institution they visit, not with an averaged machine. Their question meets your knowledge. Not an approximation.
It's a future where museums, galleries and heritage sites are not competed against by technology, but equipped by it. Where transmission keeps its seriousness and gains availability. Where depth is no longer a privilege of a few visitors.
That is the project we are working on.
Antoine and Maxime, cofounders.
Let's give your venue a voice.
Join the first museums, galleries and heritage sites partnering with us.
Or write to us: contact@getgidmi.com