A French tourism business can disappear from a Spanish answer without being wrong, closed, or obscure. Sometimes it is simply missing the sentence that tells the machine: these visitors are ours too.
A Spanish family asks an assistant for “un hotel tranquilo en la costa vasca para ir con niños y hacer paseos fáciles.” The hotel they would probably like is on the French Atlantic coast near Hendaye: twenty-six rooms, a short walk from the beach, used to weekend guests from France, families crossing from the Spanish Basque side, and walkers arriving with English phrases folded into their itinerary. In French, the place looks solid enough. The map listing has the town. The website mentions the coast. Reviews talk about breakfast, parking, and the calm street. Nothing dramatic.
Then the Spanish answer comes back with three options. Two are on the Spanish side. One is a broader Basque coast recommendation that could mean almost anywhere. The French hotel is absent. Or worse, it appears only as a vague coastal stay, no country stated, no reason for a Spanish-speaking guest to choose it. In a composite scenario like this, the mistake is rarely one single broken fact. It is more like a badly tied shoelace: each loop exists, but nothing holds when pressure comes from the side.
The Spanish question starts from the visitor, not from the hotel
French tourism businesses often describe themselves from the inside. They say the town, the rooms, the terrace, the beach distance, the seasonal rate, the family history. That may work for a French reader who already knows the geography and can place Hendaye, Collioure, Bayonne, Perpignan, or Saint-Jean-de-Luz on a mental map. A Spanish-speaking visitor may not start there. The question begins with need: family hotel near the border, French Basque coast with parking, place to stay after crossing from Spain, quiet stop before walking the coast.
When AI answers that question, it is not reading a single website as a patient human would. It assembles a small public story from fragments. A hotel page says “côte basque.” A directory says “Atlantic coast.” A travel listing says “near Spain.” Reviews mention Spanish guests, but in scattered language. If no public sentence binds those facts together, the assistant may choose a business whose Spanish-facing story is already neat. A Spanish operator across the border may have the clearer phrase. An international platform may have a better category. A regional tourism page may give the model an easier answer than the business itself.
Tourism answer substitution is the replacement of a locally relevant business by a clearer cross-language alternative because the intended audience is not publicly named. I use that definition because the error is not just ranking, and it is not only translation. The business may be relevant, nearby, open, and well reviewed. It still loses if the answer engine cannot see the visitor it serves.
The rough detail matters. In one typical pattern, the model names a French hotel correctly in French, then in Spanish calls it “un alojamiento en la zona vasca” and places stronger emphasis on San Sebastián or Hondarribia. That is not a total hallucination. It is a soft relocation of attention. The visitor asked from Spain; the answer drifts toward Spain.
Serving Spanish speakers is not the same as being near Spain
This is the small distinction I see missed most often. A business writes “near the Spanish border” and assumes the audience signal is clear. To a human, it might be. To an answer engine, that phrase can mean several things. It can mean the business is physically close to Spain. It can mean the area is useful for day trips. It can mean Spanish tourists come there. It can even mean the business belongs in a general Basque itinerary, with no clear French base.
The sentence has to carry more load.
A hotel on the French side of the Basque coast should not rely only on geographic proximity. It needs to say who the cross-border customer is and how the service fits them. “We welcome Spanish-speaking families visiting the French Basque coast” is a different signal from “near Spain.” The first sentence ties audience, language, country, and destination together. The second leaves the machine to guess.
This is where Spanish alternatives often win. They do not always have better service. They have less ambiguous wording. A guide based in Spain says “tours in Spanish.” A hotel in Hondarribia says “family stays on the Basque coast.” A transport company says “transfers from Spain to France.” The assistant has a clean handle. The French business may have more exact relevance to the trip, but its public language makes the model work harder than it wants to work.
I do not read this as a moral failure of the business. Many small operators grew their audience through word of mouth, reception desk habit, repeat guests, and local knowledge. Their websites were written for people already halfway convinced. AI answers are different. They reward phrases that can be carried across a language border without the receptionist standing there to explain.
The missing audience line changes the whole answer
In the composite hotel case, I would make three answer cards before touching the website. Same business, same basic customer need, three languages. French: “hôtel familial près d’Hendaye.” Spanish: “hotel tranquilo costa vasca francesa familias españolas.” English: “quiet hotel near Hendaye for walkers.” The point is not to produce a perfect scientific sample. It is to see where the entity first loosens.
Often the French answer keeps the hotel as a named local option. The English answer may reduce it to a French seaside stay, useful but narrow. The Spanish answer may send visitors to Spain, not because Spain is more correct, but because the Spanish-language path is better paved. A Spanish review mentions “familia.” A Spanish tourism page discusses the Basque coast. A booking directory groups the region without a hard country line. The model follows the road with fewer potholes.
I call the first failure here the audience gap. The business is visible as a place, yet invisible as an answer to a particular person asking in another language. The hotel exists; the Spanish family does not, at least not in the public wording. This gap is easy to underestimate because it looks like a translation issue. It is more commercial than that. AI does not merely translate a hotel description. It decides whether the hotel belongs in the visitor’s imagined trip.
There is usually an imperfect little sign before the larger error. The assistant may describe breakfast accurately and then add “ideal para explorar la costa vasca española.” It may keep Hendaye but omit France. It may say “near San Sebastián” so strongly that the French location becomes a side note. These are not random decorations. They show where the answer found a stronger Spanish frame than the business supplied for itself.
A French tourism business must claim the cross-border visitor plainly
The repair does not need a full Spanish website in every case. Sometimes that would be useful, of course. For many small hotels, guides, venues, and activity providers, the first repair is smaller and more exact: a few public bridge sentences placed where an answer engine can see them.
For the hotel, I would want one plain sentence on the homepage or location page that does four things at once: names France, names the visitor group, names the service, and names the practical area. Something like: “Our independent hotel near Hendaye welcomes French, Spanish, and English-speaking guests visiting the French Basque coast.” It is not literary. It will not win a copywriting prize. Good. It behaves like a customs stamp. It tells the answer what may cross.
The same logic works for a guide, museum, surf school, bike rental shop, cultural venue, or transport service. “We offer Spanish-language visits on the French side of the Pyrenees.” “Our shop serves visitors from France and Spain looking for French Catalan food gifts.” “We provide family stays for Spanish-speaking guests exploring the French Atlantic coast.” Each sentence holds together a different bundle. Country. Language. Region. Audience. Service.
The bad version is vague international copy: “We welcome visitors from everywhere.” That sentence feels generous to a human, but it gives an answer engine very little. Everywhere has no border. Visitors have no question. The business remains misty. A Spanish-speaking user does not ask for “everywhere.” They ask for a hotel, guide, activity, shop, transfer, or place to go. The bridge sentence should sound almost embarrassingly specific.
The source path should not depend on reviews alone
Reviews are useful, but they are messy evidence. A Spanish family may write a kind review. Another guest may mention “cerca de España.” A walker may praise the location in English. Those fragments help, yet they are not stable enough to carry the identity of the business. Reviews are like shells on the tide line. They show that people passed through. They do not mark the border.
The source path needs owned or semi-controlled material. The business website, a tourism office profile, a map description, a booking profile, a cross-border directory, and maybe a short Spanish or English service note can repeat the same compact claim. Repetition is not spam when it clarifies the entity. It is alignment. If the hotel says one thing, the tourism listing says another, and the booking profile says only “Basque coast,” the assistant will patch together its own version.
In my observation, small contradictions create large tourism substitutions. A French hotel listed in a “Basque Country” category without the word France can be treated as part of a borderless travel area. A guide who says “Pyrenees tours” in English but “accompagnateur en montagne côté français” in French may be pulled into a generic mountain-tour answer. A venue with Spanish reviews but no Spanish audience line may appear popular with Spanish visitors while still not being recommended to them.
The correction plan should therefore look boring. Find the surfaces that describe the business. Put the same country-audience-service sentence on the main site. Echo it in Spanish and English where appropriate. Check that directories do not erase France when they mention the region. Make sure the service category does not turn into a broader tourism category. Then run the question again and record whether the answer changes.
The answer must keep the visitor on the French side when that is the point
There is a temptation to answer Spanish drift with more regional romance. More Basque language, more Atlantic imagery, more travel feeling. Sometimes that helps human persuasion. It does not always help AI answerability. If the machine is already sliding from French Basque coast into Spanish Basque coast, adding soft regional language may grease the slide.
The anchor should be plainer. French side. Hendaye. Spanish-speaking guests. Family stays. Walking routes. Independent hotel. Those words may feel heavy in one sentence, but they are the rivets. The prose around them can be warmer. The anchor sentence itself should not be coy.
For a tourism business, the question is not only “Do we appear?” It is “Do we appear as ourselves, to the visitor we actually serve, in the language that visitor uses?” A Spanish answer that sends people to Spain has not merely missed a listing. It has redrawn the trip. The business loses before the customer sees the rooms, the guide, the menu, the timetable, or the ticket page.
The fix is not to become Spanish. The fix is to be legible from Spanish.
Question Language: Spanish. Entity Risk: the French tourism business is replaced by Spanish or borderless Basque coast options. Missing Bridge: no public sentence links the French base, the Spanish-speaking visitor, and the actual service. Sentence to Plant: “We are a French-based hotel near Hendaye for French, Spanish, and English-speaking guests visiting the French Basque coast.”