The Korean comes back, someone pastes it in, and the page is subtly wrong. Headings break in odd places, a button has grown a second line, one paragraph is set in a typeface nobody chose.
None of this is a translation problem, and none of it gets fixed by editing the Korean. It’s four mechanical things, and you can check for all of them yourself.
1. The browser will split Korean words in half
This is the one that surprises people, because it doesn’t happen in English.
Korean is written with spaces between words, so you’d expect lines to break at those spaces. By default they don’t. The browser treats each Hangul syllable block as a valid break point and will end a line in the middle of a word.
I ran a Korean sentence through a headless Chrome at every container width from 60px to 260px. With the default setting, the line broke inside a word at 16 of the 21 widths. Here’s one of them, at 60px:
팀에 필|
요한 모|
든 것을 |
하나로 |
모았습|
니다
Three of those six lines end mid-word. To a Korean reader it looks the way everyth- / ing would look to you, except there’s no hyphen to signal it.
The fix is one CSS declaration:
word-break: keep-all;
With that set, the same sweep broke inside a word at 0 of the 21 widths.
There is a trade-off, and it’s worth knowing before you apply it globally. keep-all means a word will never be split, so a box narrower than the longest word overflows instead. In the same test the longest word measured 119px; at a container width of 100px the text ran outside its box. So: set keep-all, then check that your narrowest containers — buttons, nav items, table cells, mobile cards — are wider than the longest word they’ll ever hold.
2. The text changes shape, not just length
Each Hangul syllable is a square block assembled from two to four letters — 어, 한, 국, 값. However many letters go in, the block occupies one square cell: wider than a Latin character, and carrying a whole syllable rather than one sound.
Two consequences pull in opposite directions:
- Per character, Korean sets wider. A run of Korean occupies more horizontal room than the same number of Latin characters.
- Per sentence, Korean often runs shorter. The same idea usually needs fewer words, because a lot of what English spends separate words on — articles, prepositions, auxiliaries — is carried by particles attached to the word.
Which effect wins depends on the sentence, so the useful move is not to memorise a ratio. It’s to know where the variance hurts: anything with a fixed width and no room to grow. Buttons, tabs, nav items, table headers, the label under an icon, a card title with a line clamp. A headline set to fit on two lines in English will fit on two, one, or three in Korean, and you want to find out which before launch.
3. Text inside images doesn’t translate
Any text baked into a hero image, a diagram, a chart, an app screenshot or a social card stays in English permanently. It won’t reflow, it can’t be edited by whoever writes your Korean, and it’s invisible to search — which also means it’s invisible to your translator when they audit the page.
It’s also the item with the longest lead time to fix, because it goes back to whoever holds the source files. Finding it after the copy is delivered is the standard way a Korean launch slips.
Check for it early: open the page and try to select the text with your cursor. If it doesn’t highlight, it’s an image.
4. Your brand font probably has no Hangul in it
Most Latin webfonts contain no Hangul glyphs. When the browser meets Korean text in a font that can’t render it, it doesn’t fail visibly — it silently falls back, character by character, to whatever system font can.
The result is a page where the English headline is in your brand typeface and the Korean beside it is in the operating system’s default, at a different weight and a different optical size. It reads as sloppy without reading as broken, so it survives review.
Two ways out: pick a Korean font deliberately and pair it with your Latin one, or choose a family that ships both. Either way it’s a decision, not a default — and it has to be made per weight, because a family can cover Hangul at regular and fall back at the light weight your headlines use.
Checking your own page in ten minutes
You don’t need Korean for any of this.
- Paste the delivered Korean into your page in a browser and narrow the window to 360px. Most of what breaks, breaks there first.
- Look at the ends of lines. A line that ends without a space, where the next line starts without one, is a word split in half — item 1.
- Look at every fixed-width element: buttons, nav, tabs, table headers. Anything wrapping to a second line or clipping is item 2.
- Try to select the text in every image on the page. What won’t highlight is item 3.
- Compare the letterforms of a Korean heading against a Korean paragraph. If the shapes or weights don’t match the rest of the page, you’re looking at a font fallback — item 4.
None of this asks you to judge the Korean. It’s the half of a Korean launch that’s checkable from outside the language, and it’s the half that tends to be found last.
The other notes in this set: where machine translation stops, picking one Korean voice for your whole site, and what still looks foreign after the Korean is right.
When the page itself has to change
Some of this you can fix in an afternoon — a CSS line, a wider button. Some of it means the page was drawn around English and needs redrawing around Korean, and that’s the work I do: you bring the text in every language, and I rebuild the pages so they hold it, to each market’s own conventions. Send a page and you get one exact number up front.