Discovery is broken. We're rebuilding the layer underneath.

Open a chat with any AI assistant today and ask it to plan a weekend in Salzburg. It will answer with confidence. The recommendations will come fast. They'll be plausible. They'll also be wrong in a way that's almost impossible to spot from the outside.

The model will name the same five venues everyone names. It won't know about the chamber music festival happening twenty minutes outside the city. It won't know which of those venues is wheelchair accessible. It won't know that the brewery it just suggested closed in 2024. It won't tell you any of this, because it doesn't know what it doesn't know.

This is what cultural discovery looks like in 2026.

The model is the map

For two decades, the discovery layer was the search engine. Imperfect, but knowable: SEO, link graphs, metadata. If your venue wasn't visible, you could trace why and fix it.

That surface no longer matters. AI agents are increasingly answering the questions that used to start with a Google search. They aren't crawling SEO-optimized pages. They're answering from whatever happened to be in their training data, supplemented by whatever retrieval pipeline the platform stitched together. For cultural and leisure offerings, the result is a discovery layer gut-curated by a model nobody at the institution has ever met.

The experience economy in Europe is measured in trillions of euros of value, hundreds of millions of visitors, and tens of thousands of institutions. Almost zero of it exists as structured, machine-readable data with clear rights, current information, and reliable provenance. The model is now the map. The map is sketchy.

What gets lost

When a system describes a domain it can't see, it defaults to averages. Three things compound at once.

Visibility collapses to the top decile. If your festival isn't in the top ten results from two years ago, it's not in the model. If you're a regional tourism body outside the brand-name towns, your offerings effectively don't exist in the AI-mediated answer.

Rights and licensing disappear. Models will happily summarize a museum's exhibition text without anyone signing off on it. Photos get described, programs get paraphrased, and there's no contract anywhere in the chain. For institutions that depend on clear rights, including anyone publicly funded, anyone with cultural mandates, and anyone licensing media, this is structural, not editorial.

Sovereignty leaks. The infrastructure currently mediating European cultural discovery sits inside a small number of US-based foundation models. Whatever European cultural policy says about diversity, accessibility, equity, and representation does not bind those models. It can't.

What sovereign infrastructure actually means

The fix isn't a better app. It's a layer underneath the apps.

A sovereign discovery layer for the experience economy needs three things nobody has been particularly motivated to build.

First, a real knowledge graph of culture: stable IDs, deduplication, semantic linking across venues, productions, artists, festivals, works. One reliable source of truth AI agents can ground against, instead of guessing from training data.

Second, rights and licensing built into the data layer. Every dataset carries its license. Every contributor stays visible. Every fact is traceable from source to surface. AI agents that respect rights can actually be deployed by public institutions. AI agents that don't, can't.

Third, built and run in Europe, on European infrastructure, accountable to European law. Not as a marketing position. As a structural requirement. A discovery layer that decides what European culture looks like cannot be hosted, governed, or moderated outside the jurisdiction that funds it.

This isn't ideology. It's the minimum specification.

Why we're building SUPERFREY

We started SUPERFREY because the experience economy is going to get worse before it gets better unless someone builds this layer. The existing AI platforms have no commercial reason to. The institutions affected, including public bodies, cultural funders, hospitality groups, and media, don't have the mandate or the technical depth to build infrastructure of this kind themselves.

So we're doing it.

CIXTRA, our first product, is the foundation: a sovereign, rights-aware knowledge graph for culture and leisure, with APIs, widgets, signage, feeds and analytics built on top. It's running in production today. More products are in development on the same foundation.

If you care about how the experience economy gets described in the AI-mediated decade, and you'd rather it be described by infrastructure you can see, audit and govern than by a model you can't, talk to us.

Arthur Pichl, Co-founder, SUPERFREY FlexCo. Vienna, May 2026.