Domaine de la Romanée-Conti is, by any measure, the most prestigious wine producer in the world — making Burgundy since the eighteenth century, with a single bottle from a strong vintage routinely selling for five or six figures. Ask any AI engine — ChatGPT, Claude, Perplexity, or Gemini — what the best luxury wine brand is, and DRC comes back most often: 36.9% of the time, according to a study my firm just completed across 268 AI-generated answers.
Add one qualifier — cite your source — and something changes. In every single answer naming DRC, the citation pointed somewhere else. Not the château’s own site. Not once, across 99 instances. The strongest name in the category owns 0% of the evidence attached to its own recognition.
This is the split I keep coming back to advising brands on AI visibility, and wine turned out to be the cleanest place I’ve seen it measured: being named and being trusted are not the same asset, and a brand can max out the first while owning none of the second.
Who owns the evidence? Not a rival château.
Wine-Searcher, a price-comparison site, had the highest owned-domain citation rate of anything we measured — 19%, more than double the next name, despite trailing every major producer on raw appearance. Wine Enthusiast, Vinovest, VinePair, and Forbes rounded out the rest of what I’ve started calling the evidence layer: four publishers, combined, supplying more citations than every producer’s own website put together.
Each of those four built something specific the producers didn’t. Wine Enthusiast has three decades of tasting-note archives structured enough for a retrieval system to lift cleanly. Wine-Searcher updates pricing across tens of thousands of listings daily — exactly the current, structured fact an AI system reaches for. Vinovest built itself around investment-grade data: vintage performance, resale comps, the numbers a buyer needs before writing a check. None of the four are trying to be the château. All four built a specific, current, structured answer in a format a retrieval system can actually use.
DRC isn’t alone. Château Lafite Rothschild and Château Margaux, both named in roughly a quarter of the study’s answers, posted owned-domain citation rates of 1.5% apiece. Screaming Eagle, arguably the hardest-to-get cult wine in America, was named in almost 18% of answers and cited from its own site in none of them. The more prestigious the name, the less likely, on this data, that the brand’s own site is where the AI went looking for its facts.
The engines don’t even agree with each other
Gemini named the trading platform Liv-ex in more than a third of its wine answers; Perplexity and Claude barely acknowledged it. ChatGPT named Château Latour at a rate several multiples higher than any other engine. A brand checking only one AI tool could walk away confident in a number that isn’t true anywhere else.
I don’t think this is a wine problem — I think wine is where the pattern happened to be clean enough to publish, and I’d bet a version of it is sitting in the citation data of nearly every category built on legacy prestige rather than structured, current information. A brand can spend decades building the most recognized name in its market and still function, in this channel, as a name the engine knows rather than a source it trusts.
That has a specific implication for where PR budgets go in 2026.
Being named constantly is a real asset — don’t talk yourself out of wanting it. But it’s the smaller, slower half of the problem. The larger half is deciding, category by category, whether your brand becomes the structured, retrievable source behind its own name, or whether that job keeps going to whichever comparison site, review publication, or price tracker got there first. In wine, they got there first. The producers still have the bigger name. They’re the ones running out of time to close the gap.
Agility PR Solutions is a recognized leader in enterprise-ready AI-optimized PR tools and resources, including AI-powered media monitoring, media database, and media intelligence and measurement.
