Japanese as Japanese, Vietnamese as Vietnamese: Localization Without Shortcuts

Last July, when I assigned the Japanese edition of our blog and website, I typed one very short line to my AI assistant: "Japanese comes out as Japanese, English as English, Vietnamese as Vietnamese." A few days earlier I had told it to localize deeply and leave nothing out. In August I added a sentence that is still our standard today: the way the company refers to itself and the way the prose is built must be genuinely Japanese, done professionally, not halfway.

It sounds like an ordinary quality request. It is actually a business decision, and it makes a small company like HOPE do far more work than running a translation tool and publishing the output.

Translating words versus translating atmosphere

A half-hearted translation is usually recognizable at once, even when every sentence is grammatically correct. A reader decides within seconds whether to trust a web page, and they rely on things no dictionary holds: the level of honorific language, how the company refers to itself, the length of the opening sentence, whether it speaks plainly or circles around the point.

A warm, informal Vietnamese sentence, converted word by word into Japanese, becomes either too casual or cold. An English-language reader wants the main point first, short sentences, little exclamation. A Vietnamese reader accepts, even expects, an opening with some warmth.

So for me, translation means moving the atmosphere, not moving the words. Facts and the central point stay fixed; rhythm, order and level of formality are rewritten for the reader in that place. Even forms of address must be chosen again: how a brand names itself and its reader in Vietnamese, English and Japanese has never been three translations of one word. HOPE's three editions are not one original plus two copies. They are three pieces with the same content, each written for its own reader.

Where I learned to look at it this way

I lived, studied and worked in Japan for many years, including more than seven as a bridge systems engineer (BrSE). The job teaches something textbooks rarely say: a bridge does not translate words, it makes two sides understand the same thing, including what neither side says aloud.

A very gentle request from the Japanese side can be a task that must be done. A direct sentence from a Vietnamese engineer can be heard as disrespect. It took me years to hear those layers, and since then I have stopped believing that "translating the meaning correctly" is enough.

The rest I learned somewhere unrelated to technology: part-time shifts at a Japanese restaurant, waiting tables, taking orders, handling the bill. There I understood that omotenashi is not a slogan but a chain of very small behaviors: seeing what a guest is about to need before they say a word. That spirit shows up in writing too. A website serving Japanese customers without that attentiveness gives itself away in the first line, however perfect the grammar.

AI makes this affordable, but it does not do the checking

Native-quality work in three languages used to mean a translation team, which a small company does not have. AI changes that arithmetic. Our health blog at ikihealing.com now has more than 300 articles in Vietnamese, with English and Japanese versions available.

But fast also means wrong fast. So we set two rules.

• Translations need a human reviewer, just like originals. For Japanese, the final test is a native reader.
• The banned-claims gate applies to every language edition, not only the original.

The second is the most valuable lesson. A clean Vietnamese original does not guarantee a clean translation. Every language has its own habits of overstatement. A neutral Vietnamese sentence can, in another language, sprout a promise the original never made, because the most natural phrasing in the target language happens to sound like a commitment about effects.

Our products are dietary supplements. We do not say they are medicine, we do not promise results within a set number of days, and we do not showcase customers' health outcomes. Those lines must hold in all three languages, so the gate runs separately for each edition. How we handle legal boundaries as part of the product is told in The legal guardrail is part of the product; why AI may read widely but write narrowly is in AI reads wide, writes narrow.

This is an operating principle we set ourselves, not a promise that nothing will ever slip through. That is why a person reviews, and why we keep measuring.

A door, not yet a result

When a Japanese or international partner looks into a Vietnamese company, the first thing they meet is its website. If that site speaks their language in their tone, it delivers a message no one has to say aloud: this company is serious and respects its reader.

I want to be clear that this is a door, not a result. This essay sets no expectation number for the Japanese market, and I do not think it should. What I am sure of is that if the door still looks machine-translated, a partner who wants to walk through will hesitate.

What investors and partners should understand about HOPE

• AI has sharply lowered the cost of localization, so the remaining advantage is taste and discipline: knowing what feels natural to each reader, and knowing when to stop and check.
• This is an investment in credibility, not a promised revenue channel.
• The review gate is part of the product, and it applies to every language, not only the one we write best.

I do not think any of this is remarkable. It continues an old habit: when you stand between two sides, your job is to help both understand each other, not to make things convenient for yourself. Japanese as Japanese, English as English, Vietnamese as Vietnamese. That is also how we want to be read: in the language of the person sitting across from us.