I Shipped This Blog in Korean, Then Rewrote It in English the Same Day
I built this site in Korean, which is my first language, and published three posts. Then I rewrote all of it in English before the day was out.
The obvious middle path — ship both, add a language toggle — is the one I rejected, and it took some looking to work out why it was wrong.
The thing that decided it
My first three posts were about setting up a terminal workspace manager, wiring TTS into a coding agent, and the YouTube Data API’s consent flow.
Almost nobody searches for those in Korean.
That isn’t modesty about the writing. It’s that the population of people who have hit a compinit cache problem and would look for the answer in Korean rounds to nothing. The people with that problem search in English, because that’s the language the tool’s docs, issues and error messages are already in.
I’d written for an audience that wasn’t there. The niche picked the language; I just hadn’t noticed.
Why not both, then
A toggle seemed like a free win. It isn’t, for three reasons.
1. Machine translation is explicitly a policy problem
Google’s spam policies name machine-translated content without human review as a category of scaled content abuse. Not a ranking penalty in some vague sense — a named violation, enforced since March 2024.
So the cheap version of a language toggle — run everything through translation, serve both — is the version that gets you deindexed. Any real bilingual site means writing each post twice, properly.
Which leads to:
2. The maintenance math is worse than it looks
Three posts becomes six artefacts. Every future post becomes two. That’s not a 2× cost on writing alone; it’s 2× on corrections, 2× on link updates, 2× on the “I learned this was wrong” edits that a work-log blog generates constantly.
And the two versions drift. You fix a factual error in one and forget the other. Now the site is confidently wrong in one language, which is worse than not having it.
3. A partial toggle is a broken toggle
A language switcher implies parity. Click it and land on a 404, or a stub, or the same English page you were already reading, and the site reads as unfinished. You only get to ship the toggle once everything exists in both languages — meaning you carry the full cost before you get any of the benefit.
The part I’ll say out loud
There’s a revenue dimension and I’d rather be direct about it than pretend the decision was purely editorial.
Ad rates vary enormously by where the reader is. Published 2026 figures put Australia around $39 CPM and the US around $36, while the lowest-rate markets sit under $1 — a spread of roughly 40–50×.
English doesn’t just mean more readers. It means readers in the markets advertisers pay most to reach. For a site that might carry ads eventually, writing in a language whose speakers are concentrated in a lower-rate market is a decision with a price tag, and I’d rather make that choice knowingly.
What it cost
Rewriting three posts, not translating them. Different framing, different examples, different assumptions about what the reader already knows.
The YouTube post changed the most. The Korean version walked through the four obstacles chronologically, the order I hit them. The English version leads with the one that permanently destroys your video with no appeal, because a reader who bounces after two paragraphs should still leave with the thing that matters.
That reordering wasn’t a translation decision. It’s the kind of thing that only surfaces when you rewrite rather than convert — which is an argument for rewriting beyond just staying inside the policy.
If you’re deciding this
The questions that turned out to matter:
- Where does your topic’s audience already read? Not where you’re most comfortable writing. For anything tooling-adjacent, the docs and the error messages are in English, so the searches are too
- Can you genuinely maintain two versions? Not write — maintain, through every future correction
- Is machine translation anywhere in the plan? If yes, it’s not a plan, it’s a policy violation with extra steps
Picking one language and doing it properly beat doing two badly. The site got smaller and better on the same day.