English lets you write a page and decide on tone as you go. A sentence can be warm or brisk, and the grammar doesn’t care.
Korean asks first. Every finished sentence carries an ending that sets the distance between writer and reader, and there is no neutral option — you pick one before the first sentence, and the page has to hold it.
Two of them matter for business copy.
The two
합니다체 — the formal register. Sentences end in -니다: …합니다, …입니다, …있습니다. Composed, a little distant, the voice of an announcement or a report. It’s the one I default to for a company the reader hasn’t bought from yet, because the distance reads as composure.
해요체 — the softer polite register. Sentences end in -요: …해요, …이에요, …있어요. Still polite, still perfectly appropriate for customers; the distance is just shorter. It reads like a person talking to you rather than a company addressing you.
The mistake worth heading off: these are not “formal” and “informal.” Both are polite. Korean does have genuinely casual registers, and neither of these is one. Asking a translator for “polite Korean” therefore tells them nothing — it’s true of both. Ask for 합니다체 or 해요체 by name.
Picking one
A serviceable default, and the reason for it:
| Your reader | Register | Why |
|---|---|---|
| B2B, enterprise, finance, healthcare, professional services | 합니다체 | First contact with a company you don’t know yet. Composure reads as competence. |
| Premium consumer brands | 합니다체 | Distance is part of what you’re selling. |
| Consumer apps, lifestyle, younger audiences | 해요체 | Warmth costs nothing here, and formality can read as stiff or corporate. |
If you’re unsure, 합니다체 is the safer starting point — it is rarely wrong, only sometimes cooler than you wanted. 해요체 in the wrong context is a more noticeable miss.
Two things that override the table. If your brand already has a spoken Korean voice — a support team, a community, a Korean-speaking founder — match what they already sound like; the site shouldn’t be a stranger. And if you’re in a category where your Korean competitors have visibly converged on one register, that convergence is information about what your buyer expects.
The failure is mixing, not choosing
Picking the register your reader wouldn’t have picked is a small cost. Running both on the same site is a much bigger one, and it is by far the more common outcome.
It rarely happens on purpose. It happens because a site is not written all at once:
- The landing page is transcreated properly, and the form validation messages come from the app’s string file, written a year earlier by someone else.
- Buttons and menu labels get done separately, often by whoever built the page.
- Automated email — receipts, password resets, onboarding — lives in another system entirely.
- Legal and footer text arrives from a law firm in a register of its own.
- Someone adds a new section later and matches the last thing they read rather than the site.
What I’d expect, and what I’d want checked rather than taken on my word, is that nobody reports it as a list of errors. It surfaces as a vague sense that the page is slightly off — which is harder to act on than a mistake someone can point at.
Checking it without reading Korean
You can audit this yourself, because the register lives in the last syllable of the sentence.
Pull every Korean string you ship — page copy, buttons, error messages, email templates — and look at what each finished sentence ends with:
- -니다 → 합니다체
- -요 → 해요체
Two caveats, so you don’t chase the wrong thing:
Headlines, buttons and menu labels often end in a noun with no ending at all. 배송 조회 — “track your order.” That’s not a third register; it’s the normal way Korean writes labels, the same way English headlines drop articles. Leave them out of the audit.
Requests and questions take different endings in each register, so don’t only search for the two above. In 합니다체 a request ends in -십시오; in 해요체 it ends in -세요. The -세요 form is common and soft enough that it drifts into otherwise formal sites more than anything else does — it is the single most likely place you’ll find a mix.
What you’re looking for is not a violation count. It’s whether one register is clearly the site’s and the other appears in a handful of places. That’s a list you can hand to whoever wrote them, and it doesn’t require you to judge a word of the Korean.
Put it in the brief
One line, before anyone writes:
Register: 합니다체 throughout, including buttons, form errors, and transactional email.
The clause after the comma is the one that does the work. Without it the main pages come back consistent and the edges don’t, and the edges are most of what a customer reads after they’ve bought.
The other notes in this set: where machine translation stops, what your layout does when the words become Korean, and what still looks foreign after the Korean is right.
Not sure which register your Korean is in?
Send me a page of it and I’ll tell you what’s there, and what it takes to put the whole site in one register. You get one exact number and a timeline before anything is agreed. Projects start at $400, quoted per project by the pages you want in Korean.