A multilingual chatbot you write once and answer in ninety-five languages

A multilingual chatbot is an assistant that understands and replies in more than one language. The better ones detect the language of each incoming message and respond in it automatically, rather than making the visitor pick from a menu. The distinction that matters is architectural: some are several bots in a trench coat, one per language, each needing its own content and upkeep; others keep a single knowledge base and handle the language at the moment of answering. Only the second stays maintainable past two or three markets.

Train an assistant on your existing content in whatever language you already have it, then check that a question asked in another language still retrieves the right passage and comes back correctly. Set your welcome and handover copy, and test the two or three languages that matter commercially with a native speaker before launch. What you should not do is translate your whole site first — that is the expensive step, and with this approach it is unnecessary.

Not on its own. That instruction handles composing the reply, but the harder half happens earlier: the assistant still has to find the right passage in your content, and the question arrives in a different language from the pages it is searching. If the retrieval step cannot bridge that gap it will fetch the wrong passage and then translate it beautifully, which is worse than saying nothing. Test by asking in another language and checking the fact, not the fluency.

Ours works across all supported languages on the free plan — there is no language tier and no per-language charge. One assistant and 50 conversations a month is enough to test how it handles the markets you care about before spending anything.

No. That is the point. The assistant reads your existing content in whatever language it is written and produces the answer in the visitor’s language. Translating your site is still worth doing for SEO and for the reading experience, but it is not a prerequisite for multilingual support.

From the message itself. Each message is assessed independently, so a mid-conversation switch is handled naturally and no visitor has to find a language selector.

More than 95, covering every widely spoken language and a long tail of smaller ones. The practical limit is the underlying model, and the models available in CustomerBot span the major frontier and open families — several of which are notably strong on non-European languages.

Honestly, no. Quality is best in high-resource languages — English, Spanish, French, German, Portuguese, Japanese, Chinese — and gets more variable further down the list. If a specific language is commercially important to you, test it properly with a speaker rather than assuming parity.

Yes. The system prompt is yours to write, so an instruction to always reply in a particular language regardless of input is respected. Some regulated businesses need exactly that.

No. Pricing is per conversation and per model tier, not per language. A conversation in Korean costs what the same conversation would cost in English.