Ask most of our clients if their site is multilingual and they’ll say yes. Ask them how many of those languages actually sound like they were written for that market, rather than translated into it, and the answer gets a lot less confident. That’s really the question that matters: not whether the site is multilingual, but how many languages it needs and how different those languages actually have to be.
For Swiss businesses that usually means German, French and Italian, the country’s three main languages, plus English for reach, and often the cross-border DACH markets of Germany and Austria on top.
Each of those carries traps that generic translation tools miss. Swiss German drops the ß and leans formal, using Sie rather than du. Suisse romande French isn’t France French. Ticino Italian has its own register. German-German and Austrian differ again in tone and legal wording. Treat “German” as one language, and that’s exactly how a site ends up sounding foreign in three of its own markets.
Translation and localisation are not the same job
There’s a distinction underneath all of this that quietly decides the whole architecture: translation versus localisation. Translation is the same content in another language, and it’s automatable. Localisation is different markets genuinely saying different things, and it should never be automated. Most Swiss and DACH projects mix the two, field by field.
Get the early calls right here and everything downstream gets easier. That means translated slugs and correct hreflang tags (the markup that tells Google which language version of a page to show a given searcher), fields that should never be translated at all like prices in CHF or product codes, and remembering that Swiss German simply isn’t German German. Someone needs to sign off each language before it goes live, and stale content needs to get flagged the moment its source changes.
The CMS (Content Management System) you’re on matters less here than the process does. See our companion piece, The CMS You Pick First Is Probably the Wrong One, if you haven’t settled on a platform yet. All four we build on handle multi-language content just fine. The real work is in getting the process right, not picking the tool.
The most valuable AI feature isn’t a chatbot in the sidebar
It’s the four languages that stay up to date without anyone chasing them.
Our translation pipeline runs the same way no matter which platform sits underneath it:
- Author the source. One language, written properly.
- Translate each field with DeepL. Titles, meta, slugs, alt text, not just the body copy.
- Run an LLM voice and glossary pass. This is where brand voice and an enforced glossary get held in place. Whether it’s Sie or du is a brand decision, not a linguistic one.
- Human review. Nothing auto-publishes. This step is the whole reason the other four are trusted.
- Publish per market, field by field, so the layout is never touched.
That fourth step is really the whole point. It’s why this is a service, not a plugin. On Payload builds specifically, that review gate lives directly in the content model: AI-suggested translations sit in their own fields, separate from approved copy, until someone actually signs off. More on how that structure works in how we automate content workflows with AI.
Swiss data, handled properly
AI translation means content passes through DeepL and an LLM, and Swiss decision-makers are right to ask where that data actually goes. We treat this as a first-class requirement, not an afterthought. That means processing that respects the revised Swiss Data Protection Act (nFADP, in force since September 2023) and the GDPR, Swiss or EU hosting where it’s needed, and a human approval gate that doubles as a compliance control. Nothing reaches a live market without a person signing it off.
The same step that protects your brand voice protects your data governance.
What this actually looks like on a real project
Take a client running content in German, French and Italian on one of our WordPress builds. Every new article gets authored once, in German, then runs through the pipeline above automatically: a DeepL draft, a glossary-checked LLM pass, then it sits in a review queue for French and Italian sign-off before anything publishes. Nobody’s manually retyping prices or double-checking whether “Sie” survived the translation. The system handles that.
What the human reviewer actually spends their time on is the judgment calls: does this paragraph read like it was genuinely written for a Ticino audience, or does it just sound translated for one.
Ready to get your multi-language content right?
Whether you’re on WordPress, Payload, or another CMS, or still weighing which platform fits your project, we’d love to talk. We build multi-language sites on WordPress, Payload CMS, Statamic and Craft with the same AI translation pipeline either way – take a look at our WordPress services or our Payload CMS services to see how we approach it.
What our clients ask us
FAQs
Is the AI translation nFADP- and GDPR-compliant, and can you host in Switzerland?
Yes. Content processing respects the revised Swiss Data Protection Act (nFADP) and the GDPR, we offer Swiss or EU hosting where it’s required, and nothing publishes without human approval.