AIRI-Wine industry standard · 26 July 2026 · 8 min

Country comparison: where wineries are visible to AI — and where they are not (CH · DE · AT · IT · FR)

Five countries, 100 wineries, 180 stored AI answers: Austria leads with 20% visibility, Switzerland comes last with 8% — despite a good AIRI average.

Photo: Pexels

Five countries, 100 wineries, 180 AI answers stored word for word — all measured on 26 July 2026 with the same procedure. The result turns a widespread assumption on its head: it is not the great wine nations that lead, but Austria.

Three findings

First, so the numbers are read correctly: visibility is an outcome, not a property. It comes out of three things. First, how well known you are away from your own website — Wikipedia, press, guides, retailer pages, links. Second, readability: can a machine get in at all and find verifiable facts (this is what the AIRI score measures). Third, presence: is there anything there that answers the specific question. For the world-famous names the first dominates — they get recommended regardless of how their website is built. For everyone else the second and third are the only levers they hold themselves, and they work only together: readability without presence stays invisible. The findings below should be read along that dividing line.

1. Visibility does not follow the size of the wine nation. Austria (20%) and Italy (17%) lead, France (13%) sits behind Germany (16%), and Switzerland (8%) comes last. AI systems recommend what is described often and consistently — a dense, well-documented region like the Wachau beats the sheer size of Bordeaux.

2. The AIRI score explains the middle of the field, not the top. Icons such as Romanée-Conti (AIRI 35) or Hirtzberger (AIRI 1) get recommended despite technically weak websites — their reputation sits in books, press and training data. For everyone else the relationship is clear: whoever is machine readable gets named; whoever is not does not exist for the AI. Switzerland shows the flip side: the best AIRI average (67), lowest visibility (8%) — solid websites cannot replace missing external sources, and then machine-readable facts on your own site are the only lever.

3. The cheapest measure is everywhere the least implemented. LLM instructions (llms.txt) reach no more than 29/100 in any country — in Austria 2/100. Structured data sits below 45/100 everywhere except France (49/100). These are precisely the two dimensions a winery can improve in days rather than years.

The method box shows the Swiss run as an example; each country has its own frozen question catalogue and the same three AI systems. All five country reports are published individually on the blog.