Many companies entering Japan start by translating their site. It is the obvious first step, and machine translation into Japanese is now widely used: when DeepL opened its first office outside Europe, in Tokyo in 2023, it described Japan as its second-largest market. The Japanese these tools produce is usually correct. Whether anyone in Japan searches for it is a separate question, and it largely decides whether the page is ever found.
On October 7, 2026, we tested that question directly. We took 24 English category terms of the kind that end up in title tags and headings, ran them through Google Translate, and compared how often Japanese users search for each translation with how often they search for the phrasing they actually use. For 19 of the 24, the translation was not the term Japanese searchers use. The median gap was 38 times.
The short answer
Translation and localization fail in different ways. A translation is checked against the original; a localized page is checked against the market. In search, that difference can be counted.
In our test, Google Translate produced the most-searched Japanese phrasing for 5 of 24 terms. For the other 19, the term Japanese users type was searched far more often than the translation, and 11 of the translations had ten or fewer searches a month, or none we could measure.
Our practical rule follows from that. Translate the pages people read after they have already found you: terms, documentation, specifications. Localize the pages that have to be found: category names, product pages, title tags and headings. Write the pages that have to persuade in Japanese from the start: comparisons, guides, landing pages. Google does not penalize a page for being translated. The cost of skipping localization is quieter: a correct page that targets words few people type.
Translation, localization and transcreation
The W3C’s internationalization guidance defines localization as adapting a product or its content to the language, culture and other requirements of a particular market. That covers date and number formats, currency, symbols and legal requirements as well as text, so translation is only one part of it. Transcreation goes further and rewrites the message for the new audience. The American Translators Association puts it neatly: “Transcreation is to translation what copywriting is to writing.”
For search, there is a narrower and, in our view, more useful distinction. A translation is judged by whether it means the same thing as the source. A localized page is judged by whether it uses the words the market searches with. A translation can pass the first test perfectly and fail the second, and the 24 terms below show how often that happens.
How we ran the test
We fixed the list of 24 terms before looking at any results: 12 B2B software categories and 12 consumer categories, chosen because they are the kind of phrase a foreign company puts in a page title or its navigation. No term was swapped out after we saw its numbers.
Each term went through the web version of Google Translate on October 7, 2026, logged out, using the default model. Re-running with the “classic” model gave identical output for all 24. For each term we then listed two to four Japanese candidates, the translation plus the common native alternatives, before pulling any volumes. Monthly search volumes for Japan in Japanese came from Rakko Keyword, a Japanese keyword tool whose figures are estimates built on Google Ads data, averaged over the twelve months from September 2025 to August 2026.
Where a translation returned no data, we counted it as 10, the lowest value the tool reports, so the gaps for those terms are lower bounds. Where the English could mean two things (time tracking can mean attendance or project hours), we used the reading with the smaller gap.
What the test showed
The table lists every term, B2B software first and consumer categories second. Japanese is shown as Google Translate and Japanese searchers write it, with a short gloss where the meaning differs.
| English term | Google Translate | Searches / month | What Japanese users search | Searches / month |
|---|---|---|---|---|
| e-signature | 電子署名 | 6,600 | 電子署名 (same) | 6,600 |
| expense management software | 経費管理ソフトウェア | no data | 経費精算システム (expense reimbursement system) | 1,900 |
| time tracking software | 時間追跡ソフトウェア | 10 | 勤怠管理システム (attendance management system) | 8,100 |
| customer data platform | 顧客データプラットフォーム | 20 | CDP ツール (CDP tool) | 210 |
| password manager | パスワードマネージャー | 301,000 | パスワードマネージャー (same) | 301,000 |
| project management software | プロジェクト管理ソフトウェア | 90 | プロジェクト管理ツール (project management tool) | 2,400 |
| invoicing software | 請求書発行ソフトウェア | no data | 請求書作成アプリ (invoice-making app) | 880 |
| applicant tracking system | 申請者追跡システム | 0 | 採用管理システム (recruitment management system) | 1,300 |
| web conferencing | ウェブ会議 | 590 | web会議 (same words, “web” in Latin letters) | 4,400 |
| employee engagement platform | 従業員エンゲージメントプラットフォーム | no data | エンゲージメントサーベイ (engagement survey) | 5,400 |
| help desk software | ヘルプデスク ソフトウェア | 10 | 問い合わせ管理システム (inquiry management system) | 260 |
| social media management tool | ソーシャルメディア管理ツール | 10 | SNS管理ツール (social media management tool, written with “SNS”) | 140 |
| budgeting app | 予算作成アプリ | 10 | 家計簿アプリ (household ledger app) | 60,500 |
| meal kit delivery | ミールキットの配達 | no data | ミールキット (meal kit) | 14,800 |
| running shoes for beginners | 初心者向けのランニングシューズ | 10 | ランニングシューズ 初心者 (running shoes, beginner) | 6,600 |
| skincare routine | スキンケアルーチン | no data | スキンケア 順番 (skincare, order of steps) | 27,100 |
| language learning app | 語学学習アプリ | 480 | 語学アプリ (language app) | 1,000 |
| standing desk | スタンディングデスク | 9,900 | 昇降デスク (height-adjustable desk) | 27,100 |
| noise cancelling headphones | ノイズキャンセリングヘッドフォン | 6,600 | ノイズキャンセリングヘッドフォン (same) | 6,600 |
| protein powder | プロテインパウダー | 1,300 | プロテイン (protein) | 201,000 |
| pet insurance | ペット保険 | 40,500 | ペット保険 (same) | 40,500 |
| resume template | 履歴書テンプレート | 135,000 | 履歴書テンプレート (same) | 135,000 |
| travel eSIM | トラベルeSIM | 70 | 海外eSIM (overseas eSIM) | 12,100 |
| online therapy | オンライン治療 | 90 | オンラインカウンセリング (online counseling) | 4,400 |
Five translations matched: e-signature, password manager, noise-canceling headphones, pet insurance and resume template. In each case Japanese had already settled on a direct equivalent or a loanword, so the dictionary answer and the market answer were the same. The very large figure for password manager probably reflects people looking for the one built into Chrome.
Across all 24 terms the median gap was 37.8 times: 26.4 for the B2B terms and 101.8 for the consumer terms. Sixteen of the 24 had a gap of ten times or more. Some native terms are broader than the translation (プロテイン covers far more than protein powder), but swapping in narrower alternatives such as ホエイプロテイン, whey protein, still left a median gap of 32.4 times.
Five ways a correct translation misses
The 19 misses fell into five patterns. The first tends to do the most damage and the last the least.
The concept has a different name in Japan
“Budgeting app” translates to 予算作成アプリ, an app for drawing up budgets, searched about 10 times a month. Japanese users looking for that product search 家計簿アプリ, a household-ledger app, 60,500 times. Expense management becomes 経費精算, expense reimbursement, partly because that is the task Japanese companies buy the software for. A travel eSIM is a 海外eSIM, an “overseas” eSIM: 12,100 searches against 70.
Two translations shifted meaning. “Applicant tracking system” came back as 申請者追跡システム, where 申請者 is someone applying for a permit or filling out a form, not a job applicant. It had no measurable searches, while HR teams search 採用管理システム 1,300 times a month. “Online therapy” became オンライン治療, which reads as online medical treatment; the term people use is オンラインカウンセリング. In a health category, that shift can also change how the page is read and which rules apply to it, a point we cover in YMYL supervision in Japan.
“Software” is often the wrong suffix
Japanese B2B buyers rarely search ソフトウェア, the literal word for software. They tend to search ツール (tool) or システム (system). プロジェクト管理ソフトウェア had 90 searches a month; プロジェクト管理ツール had 2,400. Help desk software came back as ヘルプデスク ソフトウェア with 10 searches, while buyers search 問い合わせ管理システム, an inquiry management system, 260 times.
Same words, different script
Japanese mixes scripts within a single phrase, and the choice can change volume. Google Translate wrote web conferencing as ウェブ会議, with “web” in katakana. Japanese users more often type web会議, with “web” in Latin letters. Same meaning, 590 searches against 4,400. We treat script variants as their own step in Japanese keyword research.
The shorter or more familiar word often wins
A loanword is not automatically the natural choice. スタンディングデスク, the katakana for standing desk, gets 9,900 searches; 昇降デスク, a height-adjustable desk, gets 27,100. Protein powder is usually プロテイン in everyday Japanese, and skincare routine is searched as スキンケア 順番, the order in which to apply products.
Sentence-shaped translations
“Running shoes for beginners” came back as a well-formed phrase, 初心者向けのランニングシューズ. People search the fragments instead: ランニングシューズ 初心者. This is probably the least serious pattern, because Google matches meaning rather than exact strings, and a page using the full phrase can still rank. We include it because sentence-shaped titles are where machine-translated pages are easiest to spot.
What Google says about translated pages
Google does not treat translation as spam on its own. Its spam policies, last updated on August 28, 2026, mention translating as one of the “automated transformations” used in scaled content abuse, where content taken from elsewhere is turned into large numbers of pages that add little. In our reading, translating your own content for a market you actually serve is a different act.
The closest thing to direct guidance came in Google’s June 2024 office hours. Asked about AI-translated pages, John Mueller said there is no special markup for declaring them, that a site should let them be indexed if it is confident in their quality and use noindex if it is not, and that “a good localization is much more than just a translation.” Gary Illyes, in the same session, said low-quality translations could affect how a site ranks and recommended review by a native speaker. Google’s documentation on localized versions adds that pages count as duplicates only if their main content remains untranslated.
So the main risk is not a penalty. It is a translated page that is indexed, reads correctly, and targets words few people in Japan type.
Do Japanese buyers notice?
The research that exists suggests they do, and more than most. CSA Research’s 2020 survey of 8,709 consumers in 29 countries found that 76% prefer to buy products with information in their own language. Among Japanese respondents the figure was 90%. Asked whether a global brand name outweighed information in their own language, Japanese respondents were the least persuaded of any country, at 50%. CSA’s earlier survey of ten non-English-speaking countries, in 2014, found that more than 70% of consumers in Japan bought only from websites in their own language.
There is a less obvious cost as well. Japan’s National Consumer Affairs Center lists unnatural Japanese among the signs of a suspicious online shop. Wording that reads like a translation does more than miss searches: to a careful Japanese buyer it can resemble the sites they have been warned about.
Which pages to translate, localize or write from scratch
Not every page needs the same treatment. In our view, the deciding question is whether a page has to be found through search, or is read by people who have already found you.
| Page | Approach | Why |
|---|---|---|
| Terms of service, privacy policy | Professional translation with legal review | The meaning must match the original, and few people find you through these pages |
| Pages Japanese law expects, such as the seller disclosure for online sales under the Act on Specified Commercial Transactions | Write for Japan | There is usually no English original to translate |
| Help center, API documentation, specifications | Machine translation with an agreed term list and native review | Readers are already users, so consistency matters more than search phrasing |
| Category names, navigation, title tags, H1s | Localize from Japanese search data | This is where the 19 misses in our test would land |
| Product and landing pages | Localize the terms, transcreate the copy | They have to be found first and then persuade |
| Guides, comparisons, articles | Write in Japanese from the start | The structure, examples and evidence should suit a Japanese reader |
Company information pages sit somewhere in between: Japanese buyers tend to expect a particular set of facts in a particular order, which we describe in the Japanese company profile page. For articles, the writing standard we work to is in writing Japanese SEO content.
A workflow that puts translation where it belongs
The test points to one step many teams skip: deciding the Japanese terms before anything is translated.
Start with a term list. For every product, category and feature name that will appear in titles, headings or navigation, write down the Japanese candidates and check the search volume and the results page for each. For most sites the list runs to dozens of terms rather than thousands, which makes it a small job next to translating the site itself.
Then translate the pages in the translation rows above, and load the agreed terms into your translation tool’s glossary so the same choice is applied everywhere; DeepL and Google Cloud Translation both support glossaries. Before anything else goes to review, have a native speaker check titles, headings, the first sentence under each heading and calls to action.
Finally, check the result in Search Console. Its query report shows the Japanese terms your pages actually appear for, and if those differ from the terms you chose, the market is telling you something. Search Console does not include Yahoo! JAPAN, so look at Yahoo! referrals in your analytics as well.
Common questions
Three questions often come up when we show this data.
Would DeepL have done better than Google Translate?
We tested Google Translate only, so we cannot say for this list. Most of the translations themselves were not wrong, though. A translation tool returns the closest equivalent in meaning, while the word a market uses for a product is a fact about that market. It has to be looked up, not translated.
Should we noindex machine-translated pages?
Google’s own answer is to noindex them if you are not confident in their quality. Our view is narrower. Pages nobody has reviewed should not be indexed, and the pages that need to be found should be reviewed and localized rather than hidden.
Our buyers read English. Do we still need to localize?
The CSA figures above suggest Japanese buyers weigh information in Japanese more heavily than buyers in most countries. Even when the final decision-maker reads English, in our work the people who research suppliers and build the shortlist search in Japanese, and that is where the terms in this article apply.
Working on Japan?
Nihongo Search is published by Amanaut, Inc., a small Japanese SEO and content company in Tokyo working with companies entering the Japanese market. The method behind the term list is in Japanese keyword research, and the full sequence is in our guide to SEO in Japan. More about how we work.
Sources
The term test is our own. Everything else comes from the sources below.
- Nihongo Search, 24-term test: Google Translate web (default and classic models), October 7, 2026; volumes from Rakko Keyword, Japan, Japanese, 12-month average September 2025 to August 2026
- Google Search Central, Spam policies for Google web search, updated August 28, 2026
- Google Search Central, English office hours, June 2024
- Google Search Central, Tell Google about localized versions of your page and Managing multi-regional and multilingual sites
- W3C Internationalization, Localization vs. Internationalization
- American Translators Association, Transcreation: translating and recreating, 2016
- CSA Research, press release on the 2020 consumer survey, July 7, 2020
- CSA Research, press release on the 2014 survey of 3,000 online shoppers, April 9, 2014
- National Consumer Affairs Center of Japan, consumer bulletin on suspicious online shops (Japanese), July 3, 2025
- DeepL, press release on its Japan office, July 3, 2023
Last reviewed: October 2026.
