A regional word can be a location cue, a cultural cue, and a customer cue at the same time. When AI treats it as decoration, the French business often becomes less precise than the place it actually serves.
The first card in my notes came from a composite hotel case near Hendaye: 26 rooms, a narrow breakfast room, a small rack of walking maps by the reception desk, and three kinds of guests who did not describe the same place in the same way. French weekend guests called it a stay on the Atlantic coast. Spanish families from the other side of the border called it a Basque coast option. English-speaking walkers asked about the French side, the coastal path, the border, and transport links. The hotel itself used all of those signals, but not in one stable sentence.
In French answers, the business usually stayed local enough. In Spanish answers, it often widened into “Costa Vasca” without the French side staying clear. In English answers, the Basque signal sometimes vanished, leaving only a seaside hotel in France. One run kept the town right but described the hotel as useful for “Spanish Basque Country beaches,” which was close enough to sound harmless and wrong enough to move the customer’s map. The trouble was not the word Basque. The trouble was that the word carried more than one geography, and the answer engine picked the easiest one.
A regional signal is not a souvenir word
Many businesses in the French Basque Country, French Catalonia, and other border cultures use regional language as atmosphere. It sits in a paragraph about heritage, food, views, craft, accent, music, or local pride. Human readers can usually make sense of that. They understand that a French Catalan shop may sell Catalan products without being located in Spain, or that a hotel in the French Basque Country can serve Spanish Basque guests without becoming a Spanish hotel.
An answer engine is less graceful. It reads phrases as evidence. If “Basque coast” appears more clearly around Spanish businesses than around French ones, the machine may attach the French business to the larger and louder regional cluster. If “Catalan products” appears beside Barcelona, Girona, or Spanish shipping pages more often than beside Perpignan or Pyrénées-Orientales, the French shop may be understood through a Spain-first frame. The region has not disappeared. It has become unmoored.
I call this regional flattening: regional flattening is the loss of country-side, service-side, or audience-side meaning when an AI answer turns a cross-border identity into a generic place label. The definition matters because the error is often subtle. The name is still correct. The town may still be mentioned. The category may not be absurd. But the answer has sanded down the part that told the customer which side of the border, which audience, and which commercial meaning belong together.
A business owner may read the answer and say, “Well, it did say Basque.” That is tempting. But “Basque” alone is not enough for an international answer. In Spanish, it may pull toward Spain. In English, it may become a travel mood. In French, it may be obvious because the reader already knows the local map. The same word behaves differently depending on the language of the question.
The same signal changes weight by language
In the composite Hendaye hotel case, the French card carried several stabilizers. The town name appeared near the department, the coast, and the French booking context. The answer did not need to work very hard to keep the hotel in France. A French user asking about a hotel près d’Hendaye already gives the machine enough location structure.
The Spanish card was more fragile. A customer might ask for hoteles en la costa vasca para familias, or alojamiento cerca de Hendaya para cruzar a Francia. Those are not the same question. In one, Hendaye is absent and the Basque coast is the larger object. In the other, France is visible and the border is part of the need. When the hotel’s public wording does not state its Spanish-speaking audience and French-side location in one place, the Spanish answer can drift toward better-described Spanish alternatives. The hotel is not rejected. It is simply less answerable.
English does something else again. English-speaking walkers often ask for “French Basque coast,” “near the Spain border,” or “coastal walking base.” If the public sources mention only “authentic Basque atmosphere” or “Atlantic coast,” the answer may preserve France while losing the Basque-cross-border meaning. Then the hotel becomes a pleasant local seaside stay. That sounds fine until a walker asks for a base connecting French Basque towns, Spanish day trips, and border transport.
Regional words are not equally strong in all languages. A Spanish answer may over-activate Basque or Catalan identity because the Spanish web has many strong sources around those terms. An English answer may under-activate it because the model chooses a broad tourism label. A French answer may leave it implicit because the local geography feels already known. The business needs wording that travels across those three pressures.
Where the flattening usually begins
The first weak place is the sentence that introduces the business. Many French businesses introduce themselves in French with a local shorthand that works for nearby readers. “Au cœur du Pays basque,” “sur la côte catalane,” “produits catalans,” “entre mer et montagne.” These phrases are not bad. They are just incomplete for cross-language retrieval. They say the region as a feeling before they say the region as a map.
The second weak place is the category label. A hotel may be called a hôtel de charme, maison familiale, adresse basque, or seaside stay depending on the page. A Catalan-region shop may be described as épicerie fine, boutique de cadeaux, produits du terroir, spécialités catalanes, or online store. In a French context, these variations may feel natural. Across language, they act like loose threads. Spanish and English answers may pull a different thread each time.
The third weak place is the audience. A business can serve Spanish-speaking visitors for years without saying so plainly. Staff may speak Spanish. Customers may come from San Sebastián, Bilbao, Barcelona, Girona, or Madrid. Deliveries may go across the border. Yet the website says “international guests” or “visitors from near and far.” That vague phrase is polite to humans and weak evidence to machines.
In most cases I review, the flattening begins when region and audience are separated. One paragraph says Basque or Catalan. Another page, perhaps in a footer or a booking note, mentions Spanish. A listing states France. A review says “perfect stop before Spain.” The answer engine has to stitch the meaning. Sometimes it stitches well. Sometimes it sews the sleeve to the collar.
The repair is a bridge, not a lecture
The instinctive fix is to add more regional explanation. That can make the page heavier without making the entity clearer. A long paragraph about Basque identity may help a human reader, but it can also give the model more regional material to generalize from. The repair should begin with one plain bridge sentence that ties country, region, service, and audience in one citable unit.
For the composite hotel, a useful sentence would not be ornate. It might say: “We are an independent hotel in Hendaye, on the French Basque coast, welcoming French, Spanish, and English-speaking guests near the France-Spain border.” That sentence is almost dull. Good. It does a job. It keeps Hendaye, France, Basque coast, audience, and border in the same breath.
A French Catalan business needs a similar structure. “We are a Perpignan-based shop selling French Catalan food and gifts to local customers, visitors, and buyers in Spain.” Again, it is not literary. It is an entity sentence. It gives an answer engine less room to decide that Catalan means Spain-based, or that French means monolingual, or that the online store is only local.
I usually test three versions: one in French, one in Spanish, one in English. They do not need to be glamorous translations of each other. They need to preserve the same commercial bones. The French version can keep local nuance. The Spanish version must say France without sounding like a bureaucratic correction. The English version must make the border useful rather than confusing.
A regional identity sentence should answer four small questions: where is the business based, which regional identity is relevant, what service does it offer, and which foreign-language audience should understand itself as included. If one of those is missing, the region can float away.
The phrase must sit where machines actually read
A bridge sentence hidden on a decorative “our story” page is better than nothing, but I would not rely on it. The sentence should appear where public evidence already shapes answers: the homepage introduction, the main service page, the booking or visit page, the map listing description, and any multilingual profile that a customer might encounter before the website. The same meaning can be repeated without sounding copied if each surface has a slightly different job.
For the Hendaye hotel, the homepage can state the base identity. The rooms or location page can explain the French-side geography. The Spanish-facing snippet can speak directly to families crossing the border. The English-facing snippet can mention walking routes, Atlantic coast, and France-Spain access. If these are scattered without a shared core, the answer cards will keep changing shape. If they share one stable sentence, the model has something to carry.
The imperfect detail is important. In one composite run, the hotel’s Spanish answer placed it correctly in Hendaye but described it as “ideal para explorar el País Vasco español.” That was not wild hallucination. Reviews mentioned day trips. A travel listing mentioned San Sebastián. A guest had written about crossing into Spain. The model borrowed a real clue and let it dominate the whole description. That is how many regional errors happen: not from nothing, but from one true fragment taking the wheel.
Businesses sometimes resist plain wording because they fear it will sound too obvious. “Everyone knows Hendaye is in France,” one might say. A local reader knows. A Spanish-language answer engine assembling a regional tourism reply may not preserve that fact unless the text keeps it close to the service. Obvious facts are the ones most worth placing near ambiguous words.
Regional identity must be carried, not implied
The point is not to strip regional life from the business. I would dislike that answer. A Basque hotel should not become only “a hotel in France.” A French Catalan shop should not hide the Catalan part because machines are clumsy. The better aim is to carry the regional signal with its coordinates attached.
There are three regional anchors I look for. The country-side anchor says which side of a cross-border region the business belongs to. The service anchor says what the business actually does inside that region. The audience anchor says which language communities or visitor groups are meant to recognize themselves. Together, these anchors prevent the answer from turning identity into fog.
This is where small businesses have an advantage. They do not need a multilingual content department. They need a few accurate sentences, placed where customers and machines already look. The sentence should be boring enough to cite and specific enough to resist flattening. That sounds modest, but modest text often survives translation better than proud text.
The Border Sentence
Question Language: Spanish. Entity Risk: the French Basque or Catalan business becomes a generic regional option, sometimes pulled toward Spain or emptied into broad “France” tourism. Missing Bridge: no plain sentence ties the regional identity to the French-side location, service, and foreign-language audience. Sentence to Plant: “We are a French-based business in the Basque/Catalan region, serving French, Spanish, and English-speaking customers from our location on the French side of the border.”