A documented AI visibility audit of one Napa Valley luxury resort shows how artificial intelligence decides which hotels a traveler considers, and where that decision goes wrong.
For many high-intent travel and event queries, artificial intelligence now assembles the first shortlist a traveler or meeting planner sees, often before they open a search engine, a booking site, or a property’s own page. The consideration set is increasingly formed inside an AI answer. That makes one question commercially decisive: when a traveler describes a place and a purpose, does the property appear in the answer?
For many luxury properties the answer is no, and not because the AI lacks the facts. Estate Yountville operates the largest event footprint in Napa Valley, 55,000 square feet across eleven venues. In the audit’s private-events query on Gemini, the property returned was Hotel Yountville, with 4,400 square feet. The larger venue was not ranked lower. It was not in the answer at all. Yet asked directly who Estate Yountville is, both ChatGPT and Gemini described it accurately. The property is known. It is simply not recommended.
That distinction is the subject of an AI visibility audit. Being known is not the same as being recommended. A property can hold every credential, sit accurately in the data, and still be absent from the answers that form demand.
Why an AI visibility audit matters now
Travel discovery is moving from a list of links a traveler scans to a single answer a machine composes. In a list, a property that ranks tenth is still visible and can still be chosen. In an AI answer, a property that is not named is not present. For many users there is no practical second page. The shortlist is the whole list.
That moves the competition for demand earlier. It no longer happens only on the booking page or in the metasearch auction. It happens in the moment the AI decides which three or four names to offer, and a property absent from that moment is excluded before price, location, or product are ever weighed.
Known by name, absent by category
The audit tests this directly. We ask ChatGPT and Gemini the questions a traveler or planner actually uses, chosen to mirror the categories that drive luxury bookings and framed by place and purpose rather than by property name. Each query is run on both platforms, and we record what each returns: the properties it names, the ones it omits, and the language it uses to describe them. Every result is captured with its platform, its exact query, and its date, then set against the property’s own verified public credentials. The findings are documented so they can be reviewed and retested as a pattern, not presented as text that will reproduce word for word on every run.
Across fifteen tested queries, Estate Yountville appeared in only one, a visibility rate below seven percent. That single appearance came when the question named the property directly and asked who it suits. In the fourteen place-and-purpose queries that resemble how travelers and planners actually search, the property did not appear.
The misses were not random. The property was absent across the queries for the best luxury resort, romantic and anniversary stays, spa and wellness, food-lover trips, weddings, corporate retreats, private events, and meeting venues. These are the categories that define its commercial value, and it surfaced in none of them.
| What a traveler or planner asks | What AI returns | What the property actually offers |
|---|---|---|
| Best hotel for food lovers in Napa | Four Seasons, Auberge du Soleil, Bardessono | The most Michelin-dense small town in the country, walking distance to The French Laundry, Bouchon, and Bouchon Bakery, with three on-site restaurants |
| Best corporate retreat venue in Napa | Carneros, Stanly Ranch, Silverado | 55,000 square feet across eleven venues, the largest event footprint in Napa Valley, including the historic Barrel Room |
| Best boutique hotel in Yountville | Bardessono, Hotel Yountville | Two distinct boutique experiences on one 22-acre campus, Vintage House and Hotel Villagio |
Every entry in the right column is confirmed and public. None of it is reaching the place where the AI composes its answer. The full set of fifteen queries and their results is in the audit.
Each omission is a specific kind of demand that never reaches the property. The event and corporate-retreat queries are group and meeting leads. The food-lover and boutique queries are high-rate leisure guests. In each case the consideration set forms without the property in it.
Why the gap exists
An AI answer is assembled from the public record the model can read. For a hotel, most of that record is not the property’s own site. It is the third-party narrative environment: online travel agencies, review platforms, destination sites, event and venue directories, and editorial coverage. The property’s own website is one voice against thousands of pages it does not control.
That material is often thin, generic, or out of date, and the pattern in the audit is consistent with the AI drawing its answer from it. The same gaps appear in both places. The booking and directory descriptions of Estate Yountville left out the property’s largest commercial asset, its event space, and named on-site restaurants that have since closed, and the AI’s answers carried the same omissions. The model was not mistaken about Estate Yountville. It was repeating the most widely distributed version of the property, and that version was incomplete. As long as the incomplete version is the most available one, it is the version the AI is most likely to return.
The gap this audit documents is not in the product. It is in how the property’s narrative is distributed across the sources the AI reads. The audit does not show that the properties winning these queries are better fits. It shows that their narratives are more present, and more consistently reinforced, in the places the model looks.
The cost of inaction
This is the point at which the finding becomes a governance question rather than a marketing one. The evidence is documented, the gap is specific, and the cause is addressable. Once that is understood inside an organization, inaction is a decision rather than an oversight.
Consider it from the position of ownership. An owner who learned that their marketing leader held this evidence, understood the property was absent from the answers shaping its own demand, and chose not to act would have difficulty seeing it as anything other than negligence. Not because a campaign was missed, but because a known and correctable leak in the property’s demand was left open. The exposure is not the audit. The exposure is holding the audit and doing nothing with it.
What closes the gap
The correction is not a better product or a new amenity. It is a corrected information environment. The objective is not to out-publish the booking platforms, which no single property can do. It is to make the property’s accurate and complete narrative present and authoritative enough across the sources an AI reads that the model has a correct version to draw on. In practice that means identifying where the AI’s account diverges from the facts, building accurate structured content about the property, corroborating it through credible external sources, and tracking whether the AI’s descriptions and recommendations change over time. The discipline that does this is Knowledge Formation Optimization. The full Estate Yountville audit, with all fifteen queries and the results behind them, is available here.
