Keyword Research for Japan: Romaji, Kanji, and What Google Actually Ranks

Keyword research for Japan breaks when teams treat Japanese like a single script. Users type スマホ, sumaho, and スマートフォン for the same device; commercial queries append おすすめ where English uses "best"; product categories split between loanwords in katakana and native kanji compounds. Google.co.jp ranks the form people actually search—not the form your translation vendor preferred. This guide covers how to run country-scoped discovery for Japan, interpret Romaji vs kanji variants in tool output, validate SERPs without fluency, and avoid the mistakes that make JP pages invisible despite "correct" Japanese.

• Target country = Japan on every discovery run—country-level volume for JP only, not global Japanese.
• Seeds in your working language (English, Chinese, etc.) are fine; output is local JP variants from Global Keyword Finder + Ahrefs metrics.
• Validate on google.co.jp with SERP structure checks—you read layout, not every character.
• おすすめ / ランキング / 比較 often signal commercial intent modifiers—pair with search intent labels.
• Seeds from support, Rakuten, Yahoo Japan—not English homepage titles alone.

1. Why Japan is not "translate and publish"

Japan is a high-value, high-friction market. Friction comes from script variation and intent phrasing, not from lacking Google data. A literal translation of robot vacuum cleaner may produce ロボット掃除機 while buyers cluster on category terms, katakana loanwords, or shorter compound queries you will not guess from English.

The classic failure: optimize for スマートフォン because it is the "full" word while スマホ carries the search volume. Same product, same language, wrong string. That is localization vs translation at the character level—grammar can be perfect and SEO still fails.

2. Romaji, kanji, katakana: what shows up in search

You do not need to master Japanese linguistics—you need to recognize which script wins in tools and SERPs for your category.

• Kanji compounds — Common for established categories (掃除機, 自動車). Often the head noun in informational and commercial queries.
• Katakana loanwords — Imported product types and brand-adjacent categories (ロボット, スマホ, ノート PC). Dominant in consumer electronics and fashion.
• Hiragana — More common in long-tail, conversational, or younger-audience phrasing; appears mixed in suggestions.
• Romaji — Lower volume for most B2C categories but appears in tool exports and some tech queries; still validate—never assume romaji is safe because volume looks small.
• Mixed strings — ロボット掃除機 おすすめ, コードレス掃除機 ランキング; modifiers are part of the keyword, not optional ad copy.

Discovery tools return the strings Ahrefs sees for Japan—often kanji/katakana dominant. Romaji variants may appear as separate rows with different volume; treat each row as a hypothesis until SERP validation confirms which form Google surfaces in titles.

3. Intent modifiers Japanese buyers actually use

English teams map "best" and "review" to superlatives. Japanese commercial research uses different surface forms—visible in SERP titles without reading every word.

• おすすめ (osusume) — "Recommended" — comparison and roundup SERPs; high commercial value.
• ランキング — Ranking lists; often product-grid style results.
• 比較 — Compare; good for category hubs.
• 口コミ — Word-of-mouth/reviews; informational-leaning but critical for trust categories.
• 安い / 価格 — Cheap / price; transactional undertone—check SERP for shopping modules.

When you run discovery from an English seed like robot vacuum, compare whether JP variants attach おすすめ or category nouns—not whether the English "best" was translated. See non-English markets playbook for the cross-market intent pattern.

4. Workflow: Japan discovery pass

One country, one sprint—same unit as Germany in the DE/AT/CH guide, but script-aware validation.

• Step 1 — Seeds (15–25): English or Chinese product/problem phrases; add strings scraped from Rakuten, Amazon.co.jp, Kakaku.com category leaders, Yahoo!知恵袋 question titles.
• Step 2 — Discover: Global Keyword Finder → Source language = your sheet; Target country = Japan. Export CSV with volume, KD, CPC, intent.
• Step 3 — Group variants: Cluster rows that differ only by script (スマホ vs sumaho vs スマートフォン)—pick winner by JP volume + SERP, not row count.
• Step 4 — SERP validate: google.co.jp incognito; pass if titles use same script/form; fail if Wikipedia/news dominates for a commercial seed.
• Step 5 — Page type map: おすすめ/ランキング → comparison or hub; product nouns → PLP/PDP—intent guide.
• Step 6 — Native review: Final titles and body copy—structure validation can be non-fluent; customer-facing copy needs JP reviewer.

Credit planning: 20 seeds × 1 country = 20 credits per wave. Refresh quarterly or before major SKU launches—see pay-per-search pricing.

5. SERP validation without reading Japanese

You validate patterns, not prose—same method as keyword research without fluency.

Green lights on google.co.jp: shopping boxes, Rakuten/Amazon.co.jp listings, comparison sites with numbered lists, brand PLPs, Kakaku-style price comparison. Yellow: mix of blogs and products—split query or match dominant above-fold type. Red: Wikipedia, news, government (.go.jp), Q&A only—deprioritize for product-led pages unless content strategy owns the query.

Script check: if top three titles use katakana/kanji and your page is romaji-only in the title tag, fail—even if romaji row showed some volume. Align visible URL slug policy with validated script where possible.

6. Common mistakes in Japan keyword research

These appear in JP launches that indexed but never earned clicks.

• Dictionary translation as keyword list — Fix: JP discovery pass; trust local variants over vendor output.
• Optimizing full form over abbreviation — Fix: compare スマートフォン vs スマホ-style pairs in export.
• Ignoring おすすめ constructions — Fix: commercial seeds include recommendation modifiers.
• Global Japanese volume — Fix: JP country scope only.
• Romaji URLs for katakana SERPs — Fix: validate script in titles; slug policy with dev + native SEO.
• Skipping Yahoo Japan / Rakuten seed sources — Fix: marketplace seeds.

FAQ

Should I enter seeds in Japanese or English?

English (or Chinese) seeds work for discovery when target country is Japan—the tool returns JP variants. If you have JP seeds from GSC or Rakuten, run those too and merge—two seed languages can improve coverage.

Does Yahoo Japan still matter?

For seed mining and Q&A phrasing, yes—知恵袋 questions reveal problem language. For ranking, Google still drives most organic SEO decisions; validate on google.co.jp.

How does KD for Japan compare to English?

KD is relative to the JP link graph in country-scoped tools—often lower than US equivalents for the same concept, but always open the SERP. A low KD with Rakuten and Amazon.co.jp locked in top five is not "easy."

Do I need hreflang for Japan?

If you have a dedicated JP URL set (`ja-JP`), yes—after keywords validate—see hreflang order. Keywords first; tags second.

Can Global Keyword Finder replace a native SEO consultant?

No. It accelerates discovery and country metrics—the step non-Japanese teams skip. Native review still owns tone, compliance, and final copy. GKF is not rank tracking or site audit.

Conclusion

Japan keyword research is choosing the script and modifier Japanese users type on google.co.jp—not publishing the translation your CMS exported. Run country-scoped discovery, cluster script variants, validate SERP layout, then hand winners to native writers.

Pick one product category. Run one English seed on Global Keyword Finder with target country Japan—5 free credits on registration—and open the top three JP variants on google.co.jp. If the SERP titles do not match your planned page title script, fix the keyword before you write the page.