The pattern is familiar. A company launches its Japanese site, the translation is accurate, traffic from Japan arrives, and the Japanese version converts noticeably worse than the original. Nobody can point to a single broken thing.
Usually nothing is broken. The words were translated; the site was not localized. Japanese users arrive with expectations about how much a page explains, where company information lives, what a form asks for and how a price is shown. A site that ignores them does not look wrong so much as foreign, and foreign reads as risk.
Translation is not localization
Translation converts text; localization converts the experience. Three expectations do most of the damage when a literal translation ignores them:
- Information density. Japanese product pages tend to be long and specific: specs, pricing detail, case studies, FAQs. A minimal landing page with one headline and one CTA can read as if something is being withheld. Keep the whitespace; add depth below the fold.
- Trust signals. Buyers, B2B buyers especially, check who you are before they commit. Customer logos, case studies (導入事例), certifications such as the PrivacyMark (Pマーク) or ISMS, and a Japanese support contact with stated hours all carry weight.
- Explicit process. Say what happens next: whether sales will call, when a trial ends, how cancellation works.
Microcopy: before and after
Tone is where literal translation shows most. English product copy is casual and imperative; good Japanese copy is polite (です・ます style) and specific.
| Context | Source (EN) | Literal | Localized |
|---|---|---|---|
| B2B primary CTA | Book a demo | デモを予約 | 資料ダウンロード（無料） beside お問い合わせ |
| Error | Oops! Something went wrong. | おっと！問題が発生しました。 | 申し訳ございません。エラーが発生しました。時間をおいて再度お試しください。 |
| Email greeting | Hi Taro, | こんにちは、太郎さん | 山田 太郎 様 |
| Form button | Submit | 送信 | 入力内容を確認する |
The CTA row matters most. In Japanese B2B, downloading a service brochure (資料) is a common first step, because the buyer can circulate it internally before talking to sales. "Book a demo" alone skips that step.
IA and navigation: use the labels people scan for
Japanese corporate sites share a strong navigation vocabulary, and users scan for those exact words.
| Purpose | Conventional label | Notes |
|---|---|---|
| Company | 会社情報 / 企業情報 | Hub for 会社概要 (overview), 沿革 (history), アクセス (directions) |
| Products | 製品情報 / サービス | Match your business model |
| Investors | IR情報 | Listed companies |
| Careers | 採用情報 | Often split into 新卒採用 and キャリア採用 |
| Contact | お問い合わせ | Usually a persistent header button |
| FAQ | よくある質問 | Link it from pricing and forms too |
| Privacy | プライバシーポリシー | Sometimes 個人情報保護方針 |
| Sitemap | サイトマップ | An HTML page, linked from the footer |
For the structural side, see GNB and footer structure best practices. Once the Japanese version exists, audit it against the source site instead of trusting the translation spreadsheet: DeepAnalyzer crawls both and exports an IA Excel sheet with Screen IDs and labels in the detected language (決済, カート, 採用情報), so a Japanese reviewer can read it without a glossary.
The company page is a conversion page
On many Western sites, "About" is a brand story. In Japan, 会社概要 is a fact sheet, and buyers read it: company name, date established (設立), capital (資本金), representative (代表者), address, headcount and business description. A foreign company without a Japanese entity should state plainly who the contracting party is and how Japanese customers are supported.
Keep the HTML sitemap page
The human-readable サイトマップ page has faded from Western sites but remains a common footer link in Japan. It helps visitors who navigate by overview and gives crawlers a flat path to every section; generate it from your IA.
E-commerce: 特定商取引法に基づく表記
If you sell online to consumers in Japan, the Act on Specified Commercial Transactions (特定商取引法) requires mail-order sellers, online stores included, to disclose details such as the seller's legal name, address and phone number, prices and fees, payment and delivery terms, and return conditions. The standard implementation is a page titled 特定商取引法に基づく表記, linked from the footer and checkout; recent amendments also cover what the final order-confirmation screen must show. Treat it as a legal review item, not a copywriting task.
Forms: where conversions quietly leak
Forms built for Western names and addresses fail in Japan in small, compounding ways.
| Field | Western default | Japanese convention | Implementation |
|---|---|---|---|
| Name | First, Last | 姓 (family) → 名 (given) | Two fields, family name first |
| Reading | None | フリガナ (セイ / メイ) | Kanji names have several readings |
| Postal code | After city | 〒 first, 7 digits (150-0001) | Auto-fill prefecture and city |
| Address | Street → city → state | 都道府県 → 市区町村 → 番地 → 建物名 | Largest unit first; separate building field |
| Phone | One format | 03-1234-5678, 090-1234-5678 | Accept with or without hyphens |
The postal code comes first for a reason: seven digits resolve the prefecture and city, so the user adds only the block number and building.
Normalize full-width input instead of rejecting it
Japanese is typed through an input method editor, and users routinely enter full-width digits (１２３) without noticing. The worst response is the classic error "半角数字で入力してください" (enter half-width digits). Accept anything and normalize:
// Postal codes and phone numbers only. Never map "ー" in name
// fields: it is a real katakana character (as in ヨーコ).
const normalizeJaNumber = (value) =>
value
.normalize("NFKC") // １５０－０００１ → 150-0001
.replace(/[‐‑‒–—―−ー]/g, "-") // hyphen look-alikes → "-"
.replace(/\s+/g, "");
NFKC also converts half-width katakana (ｶﾀｶﾅ) to full-width, which is what a フリガナ field should store. Two more conventions: forms commonly run input → confirm → complete (入力 → 確認 → 完了), and emails address the customer by full name with the honorific 様, never by first name alone.
Payments, prices and dates
Offer the methods people actually use
Card-only checkout loses Japanese customers. A realistic set:
- Credit cards, including JCB, the domestic brand.
- Konbini payment (コンビニ払い). The customer pays in cash at a convenience store with a payment number; the order ships once payment is confirmed.
- QR code payments such as PayPay, 楽天ペイ and d払い.
- Bank transfer (銀行振込), common for higher-value and B2B orders.
- Cash on delivery (代金引換) and pay-later (後払い), still expected in parts of the market.
B2B buyers often expect quotes (見積書) and invoice payment (請求書払い); if you invoice Japanese companies, check how the qualified invoice system (インボイス制度) applies.
Show tax-inclusive prices
Since April 2021, businesses showing prices to consumers in Japan must display the tax-inclusive total (総額表示). Write 11,000円（税込）, optionally followed by the pre-tax amount, and never lead with a pre-tax figure that makes the total look secondary. B2B pricing is generally outside the rule, but label 税込 or 税抜 explicitly either way.
Dates and the Japanese era
The standard format is 2026年9月23日, often with the weekday: 2026年9月23日（水）. Numeric dates are year-first: 2026/09/23. The era calendar (令和8年 is 2026) still appears in government documents and some official forms; a marketing site rarely needs it.
const d = new Date("2026-09-23T12:00:00+09:00");
new Intl.DateTimeFormat("ja-JP", { dateStyle: "long" }).format(d);
// "2026年9月23日"
new Intl.DateTimeFormat("ja-JP-u-ca-japanese", { dateStyle: "long" }).format(d);
// "令和8年9月23日"
Also announce shipping and support schedules for Golden Week, Obon and the New Year break (年末年始) in advance.
Typography: Japanese has no spaces
Japanese has no spaces between words, so a line can break almost anywhere.
Line-breaking rules (禁則処理, kinsoku shori). Closing brackets and punctuation (、。」) must not begin a line, and opening brackets must not end one. Browsers apply basic rules by default; line-break: strict adds the stricter set, keeping small kana (ゃ, っ) and the long-vowel mark (ー) off the start of a line.
Phrase-aware headings. Mid-word breaks are normal in body text but look careless in a large heading. Chromium browsers support word-break: auto-phrase, which breaks at phrase boundaries using the BudouX model, for content marked lang="ja". Other browsers ignore it, so it is safe to ship.
html:lang(ja) {
font-family: "Hiragino Sans", "Hiragino Kaku Gothic ProN",
"Noto Sans JP", "Yu Gothic", YuGothic, Meiryo, sans-serif;
line-height: 1.8;
line-break: strict;
}
h1:lang(ja), h2:lang(ja) {
word-break: auto-phrase; /* Chromium; requires lang="ja" */
text-wrap: balance;
}
em:lang(ja) {
font-style: normal; /* Japanese fonts have no italic */
font-weight: 700;
}
- Font stack. Browsers download a web font only when it is used, so listing Hiragino first means Apple devices never fetch Noto Sans JP. On Windows, regular-weight Yu Gothic can render thin; many teams specify
"Yu Gothic Medium". - Font size on the wire. A Japanese font contains thousands of glyphs, and one weight can run to several megabytes. Subset it or use
unicode-rangeslicing (Google Fonts serves Noto Sans JP in many small slices), and stop at two weights. - Line height. Kanji are dense; body text reads best around 1.7–1.9.
- No faux italics. Japanese typefaces have no italic, so
font-style: italicproduces a synthetic slant. Use weight or emphasis dots (text-emphasis: filled sesame).
Technical SEO for the Japanese version
Choose the structure: /ja/ or .jp
| Option | Example | Trade-off |
|---|---|---|
| Subdirectory | example.com/ja/ |
Authority stays on one domain; one site to run. The right default. |
| ccTLD | example.jp, example.co.jp |
Strong local signal and trust; .jp generally requires a local address, .co.jp a Japanese company; authority starts from zero. |
Choose a Japanese domain only with a local entity and a long-term local brand. See URL structure best practices.
Declare the language properly
Connect every version with reciprocal hreflang, including a self-reference and x-default. Plain ja is sufficient because Japanese is used almost entirely in one country; ja-JP is also valid. jp is a country code and voids the annotation. Full rules are in the hreflang guide.
<html lang="ja">
<link rel="alternate" hreflang="ja" href="https://example.com/ja/pricing/" />
<link rel="alternate" hreflang="ko" href="https://example.com/ko/pricing/" />
<link rel="alternate" hreflang="en" href="https://example.com/en/pricing/" />
<link rel="alternate" hreflang="x-default" href="https://example.com/en/pricing/" />
lang="ja" also tells the browser to use Japanese glyph forms for kanji. Without it, some systems render shared Han characters with a Chinese font, which Japanese readers spot immediately.
Skip forced IP redirects, keep slugs ASCII
Redirecting by IP traps users who want another version and can hide Japanese pages from crawlers, which mostly crawl from outside Japan. Offer a dismissible language banner instead.
Japanese slugs look friendly in the address bar but are percent-encoded when shared: /ja/料金プラン/ becomes /ja/%E6%96%99%E9%87%91%E3%83%97%E3%83%A9%E3%83%B3/. Use the same ASCII slug in every locale (/ko/pricing/, /ja/pricing/); hreflang mapping becomes trivial too.
Metadata, Yahoo! JAPAN and OGP
- Titles and descriptions. Snippets truncate by pixel width, and full-width characters are wide. As a rough guide, keep titles to about 30–35 characters and descriptions to around 120, key terms first. Write them natively.
- Yahoo! JAPAN. Its web search results are powered by Google's search technology, so one Google-focused effort covers both.
- OGP. Links are shared heavily on LINE and X, so the preview card is often the first impression. Set
og:title,og:description, a readable Japaneseog:image,og:locale(ja_JP) andtwitter:card(summary_large_image). On a JavaScript app, confirm this metadata is in the server response; comparing a plain crawl with a Strong Mode (headless browser) crawl shows what non-rendering bots miss.
Launch checklist
- GNB and footer use conventional labels (会社情報, 製品情報, 採用情報, お問い合わせ)
- 会社概要 lists date established, capital, representative, address and headcount
- An HTML サイトマップ page is linked from the footer
- E-commerce: 特定商取引法に基づく表記 linked from footer and checkout, reviewed by counsel
- Forms: 姓 → 名, フリガナ, postal-code auto-fill, full-width input normalized
- Konbini, QR code and bank transfer offered where relevant
- Every consumer price is tax-inclusive (総額表示)
lang="ja",line-break: strict, phrase-aware headings, subset fonts- Reciprocal hreflang with
x-default; no forced IP redirects; ASCII slugs - Native Japanese titles, descriptions and OGP images
- Microcopy rewritten by a native writer, not just proofread
FAQ
Is machine translation good enough for a Japanese website?
It can produce a usable first draft, not a launch-ready site. The gaps are tone, microcopy and content Japanese users expect but the source never had. Budget for a native editor to rewrite key pages, CTAs and forms.
Should we use a /ja/ subdirectory or a .jp domain?
For most companies, /ja/ is the better default: authority stays on one domain and there is one site to operate. A .jp or .co.jp domain makes sense with a Japanese entity and a long-term local brand.
Should the hreflang value be ja or ja-JP?
Either is valid; plain ja is usually sufficient because Japanese is used almost entirely in Japan. What matters more is reciprocity: every version lists every other version, including itself. Never write jp, a country code.
Do we need a 特定商取引法に基づく表記 page?
If you sell goods or services online to consumers in Japan, the Act on Specified Commercial Transactions generally requires you to disclose seller and transaction details, and a dedicated page is the standard way to do it. Purely B2B or non-transactional sites are usually in a different position. Confirm your obligations with a Japanese legal advisor.
Do we need separate SEO work for Yahoo! JAPAN?
No. Yahoo! JAPAN's search results are powered by Google's search technology, so work done for Google carries over. Invest instead in native Japanese metadata and OGP that looks good on LINE and X.