Your Vibe Project Deserves i18n: A Practical Path From Hardcoded Strings to Multilingual Launch

When i18n is worth it, and when to skip it

The verdict first: if you want to launch on Product Hunt, capture English-language traffic, or get forked around the world — do it. A project introduced in English has several times the reach of a Chinese-only one. A purely Chinese utility, a personal productivity script, an MVP still validating — skip it for now, and revisit when someone asks "is there an English version?"

There's exactly one decision criterion: is the English-speaking world in your user acquisition channels? If the answer is "planning to be," then the earlier you do i18n the cheaper it is — extracting strings at 5,000 lines of code is an afternoon's work; at 50,000 lines it's a campaign.

Step one is always: have the AI extract hardcoded strings into keys

The first step of i18n isn't translation — it's pulling every hardcoded Chinese string scattered through the codebase out into keys. The AI does this fast and well. Give it an explicit prompt: "Scan the whole project and replace all user-visible hardcoded Chinese strings with i18n keys. Don't touch logic, don't touch Chinese in comments."

Key naming convention: module.location.meaning, e.g. checkout.coupon.title, settings.profile.saveButton. Never text1 or msg_abc — the key is documentation; three months from now the key alone should tell you where it appears.

One iron rule: one key maps to exactly one meaning. The word for "confirm" in a dialog and in a form may look identical in Chinese, but don't reuse one key — in English they might be "OK" and "Confirm," and a shared key makes them inseparable.

Choosing a library, in one sentence: follow your framework

Next.js → next-intl, React/Vite → react-i18next, Vue → vue-i18n. Don't agonize over comparisons; following the framework's ecosystem is the optimal choice. Put language files at locales/zh.json and locales/en.json, with structure mirroring the dot-separated key names.

URL strategy, also in one sentence: distinguish languages by path (/zh/pricing, /en/pricing) — best for SEO and most intuitive for users.

AI drafts the translation, humans review: never skip step two

The workflow has three steps: AI drafts → you (or a native speaker) review product copy → ship, then keep fixing from feedback. AI is already good enough at UI copy, but two categories must never go live straight from machine translation: legal terms and payment-related copy. "Subscribe" and "purchase" carry completely different legal meanings in English — get that wrong and it genuinely hurts.

When reviewing, focus on three spots: button labels (English usually runs longer than Chinese — does it break the layout?), empty states and error messages (where AI is sloppiest), and marketing copy (directly translated Chinese marketing reads like a scam text in English).

A money-saving trick: have the AI translate first, then have another AI (or the same one with a different prompt) role-play a native English user and tear the translated UI copy apart. Machine cross-review catches 80% of awkward phrasing; the remaining 20% needs a human.

Four traps, each with real blood on it

Trap one: plural rules. Chinese has no singular/plural; English does. Don't lazily write {n} message(s) — use your i18n library's pluralization. Arabic has six plural forms; hardcode one s and you'll hear about it.

Trap two: date, currency, and number formats. Is 2026/10/06 October 6 or June 10 to an American? Use Intl.DateTimeFormat and Intl.NumberFormat — never concatenate these strings yourself.

Trap three: string concatenation."你有" + n + "条消息" works in Chinese; in English the word order is "You have n messages" — concatenation shatters into three pieces in the target language. Use full sentences with interpolation: t('inbox.count', {n}).

Trap four: text baked into images. Chinese text in logos, screenshots, and tutorial graphics stays put when you switch languages. Audit every static asset: make localized versions where needed, swap to pure icons where you can.

Acceptance checklist: switch languages and click through the core flow

Pre-launch acceptance is simple: switch to English and click through the entire core flow — signup, core features, payment, settings. The checklist has four items: any untranslated Chinese left, any long copy breaking the layout, any sentences broken by concatenation, any 404s (missing keys).

A final judgment call: i18n is one of the highest-ROI growth moves a vibe project can make — AI has driven the extract → translate → review pipeline cost to nearly zero, while the reach of an English introduction is very real. An afternoon of investment buys you users from everywhere. Worth it.