The old advice sounded obviously correct: write for your customer. Papiamento-speaking market, publish in Papiamento. Dutch market, publish in Dutch. Tourists, publish in English.
That advice assumed your customer would search, scan a list of results, and click. Increasingly, they don’t. They ask a question, and a machine answers it.
That changes what language is for.
The intermediary is now the audience
When someone asks an AI assistant to recommend a dive shop in Curaçao or a caterer in Paramaribo, it doesn’t browse. It synthesizes an answer from what it has absorbed and can retrieve.
You aren’t being read at that moment. You’re being summarized or omitted.
So, the question isn’t only what language your customer reads. It’s what language the system answering your customer reads, and whether you appear in it. Those aren’t always the same. A Dutch-speaking customer in Willemstad can ask in Dutch and receive an answer assembled largely from English sources, simply because that’s where the material exists.
The thin-pool problem
There isn’t much content about our region, in any language. Content about Amsterdam or Barcelona is effectively infinite. Content about Oranjestad, Willemstad or Paramaribo is not.
That’s usually framed as a disadvantage. It’s better understood as an unusually cheap opportunity. When the pool is shallow, each contribution carries disproportionate weight. A single substantive page on a specific local topic can become a meaningful share of everything a model has to work with. In a saturated market that page is invisible. Here it can be foundational.
A usable framework
English is the retrieval language, even when it isn’t the customer language. The English pool is deepest, so English material is disproportionately likely to be drawn on, including for answers ultimately delivered in another language. If you publish in one language for discovery, this is usually it.
Papiamento and Sranan Tongo are trust assets, not discovery assets. They rarely surface you in an AI answer. They prove you belong to the place once someone arrives. Treat them as a conversion layer.
Where to start
- One substantive English page per core topic. Concrete and specific, not generic marketing copy. Your discovery layer.
- Local-language versions of your highest-converting pages. Your conversion layer.
- Dutch where the commercial case justifies it.
- Specificity over volume. “Diving in the Caribbean” competes with the entire internet. “Shore diving conditions at a named site in October” competes with almost nothing and is closer to what people actually ask.
Most advice about AI search is written for markets drowning in content. Ours is the opposite: shallow pool, few businesses deliberately filling it.
That won’t stay true. The businesses that become the substantive source on their subject will be the ones future answers are built from. The rest will wonder why the machine never mentions them.