Japanese keyword research: capture every variant, not just one

Keyword research in Japanese is a different job

In English you research a handful of head terms and their variants. In Japanese, the *same* intent fractures across scripts: kanji, hiragana, katakana and the English loanword all coexist for a single concept, and demand splits unevenly between them. Research one spelling and you can miss most of the real volume — and, worse, build your whole page around the form nobody actually types.

The reason is structural, not stylistic. Japanese has no spaces between words, so search engines tokenize sentences into morphemes before matching them to a query. That means the *script* a user chooses changes the token, the token changes the match, and the match changes the SERP. Three "spellings" of one idea can return three different sets of competitors.

Why one concept becomes four keywords

Take "smartphone." A Japanese searcher might type スマホ (clipped katakana), スマートフォン (the full loanword), smartphone in English, or — for an older or more formal audience — 携帯 (a kanji term that technically means "mobile phone" but is used loosely). Each form has its own volume, its own intent, and its own ranking page.

A travel brand once asked us why a beautifully optimized page for 宿泊 ("lodging," kanji) was flat. The answer was that their audience searched ホテル (katakana, "hotel") and 宿 (a shorter kanji term) far more often. The intent was identical; the demand had simply pooled on two spellings they had never checked.

Map the cluster before you write

Before drafting anything, build a variant map for each concept you want to own:

• Enumerate the kanji, hiragana, katakana and loanword forms — plus common abbreviations and any 当て字 (creative spellings).
• Pull search volume from Japan, not a global or English average.
• Read the SERP intent per variant; a single concept can mix informational and commercial results across its forms.
• Decide which one variant the page targets in its title and H1, and which others it supports in body copy and headings.

Long, conversational queries win

Japanese searchers type full questions far more than English users do, partly because IME input makes longer strings cheap and partly because spoken Japanese is naturally polite and complete. The modifiers that signal intent are predictable, and they are where the commercial and informational value actually sits.

These four suffixes routinely account for the bulk of a concept's long-tail demand. A page titled スマホ おすすめ 2026 will pull buyers; スマホ 比較 will pull comparison-shoppers one click from converting; スマホ とは will pull the top of the funnel. Treat each suffix as a distinct content brief, not a keyword to stuff into one article.

A quick worked example

Say you sell project-management software. The English instinct is to target "project management tool." In Japanese the demand fractures like this, and the right move is a hub-and-spoke set of pages rather than one:

• プロジェクト管理 ツール — head term, high competition, brand-led SERP.
• タスク管理 アプリ — adjacent intent (task not project), often higher conversion.
• ガントチャート 無料 — feature + "free," strong commercial signal.
• プロジェクト管理 とは — informational, easy to rank, feeds the cluster.

Targeting only the head term leaves the three spellings that actually convert on the table.

How demand splits between variants

When you finally pull Japan volume for a cluster, the share almost never lands where intuition says. For a consumer concept the casual katakana form usually dominates; the formal spelling and the English loanword trail; the kanji term can be a long-tail surprise. A representative split for the "smartphone" cluster looks like this:

That distribution decides everything downstream. You write the title and meta description in スマホ, you weave スマートフォン into a spec section so you still match formal queries, and you don't waste an H2 on the English form that holds four percent. Get the proportions wrong and you can rank first for a word that drives a fraction of the clicks you assumed.

Use competitor SERPs as a variant cheat-sheet

The fastest way to find variants you missed is to read who already ranks. Pull the top ten results for your head term, then for each ranking page note which spelling sits in its title, which appears in its headings, and which long-tail suffixes it targets. Patterns emerge quickly: if four of the top ten put タスク管理 in the title while you have been writing プロジェクト管理, the market has told you which token it rewards. Reverse-engineering the SERP is faster and more honest than guessing, because it reflects what Google has already validated for that exact query in that exact location.

Localize the measurement, not just the words

Volume and SERP composition both shift by location. A query that looks competitive nationally can be wide open in 大阪 or 福岡, and a local-intent term (美容院 + a ward name) only makes sense measured from the right city. Sizing a Japanese opportunity off a global figure, or off a SERP pulled from the wrong country, is how teams ship pages that never had a chance.

The practical checklist:

1. 1Build the variant map (kanji / kana / loanword / abbreviation) for every concept.
2. 2Pull Japan-specific volume for each form and note the gaps.
3. 3Inspect the live SERP per variant from a Japanese location to confirm intent.
4. 4Assign one primary variant per page; support the rest in headings.
5. 5Track all variants over time — share between spellings drifts as language and trends move.

Measure it where it matters

Lume runs Japanese keyword research against real local volume and location-accurate SERPs across 100k+ geographies, so you capture the variant that actually converts instead of the one an English tool surfaced. Run a free scan and see your true Japanese demand.