Guide · Japanese

Japanese pages in search and AI answers: real data from two sites

Updated 2026-11-12 · kagentx

Short answer

In MochiStamp, Japanese pages had impressions on 34 pages in the last 28 days (Google Search Console, 2026-10-14 to 11-10): 833 impressions, 19 clicks, average position 9.6. In Gamma AI Slide, 8 /ja/ pages got 56 impressions and 4 clicks at position 6.0. The pages with impressions almost all target a specific question Japanese users search for (specs, descriptions, phrases). The sample is two sites and four weeks, and the dashboard cannot confirm whether Japanese pages are cited in AI answers.

How do Japanese pages compare with other languages on impressions, clicks and position?

We have two sites with Japanese pages: MochiStamp makes LINE stickers, Japanese is its main language and its pages have no language prefix; Gamma AI Slide makes presentations and its Japanese is under /ja/. Below are the last 28 days (Google Search Console, 2026-10-14 to 11-10), added up by language prefix.

Site and languagePages with impressionsImpressionsClicksClick-through rateAverage position
MochiStamp Japanese (no prefix)34833192.3%9.6
MochiStamp English2667381.2%8.9
MochiStamp Traditional Chinese30637284.4%8.2
Gamma AI Slide Japanese (/ja/)85647.1%6.0
Gamma AI Slide English (no prefix)3344910.2%27.4

Average position is weighted by impressions, the "other" bucket with no URL is not counted, and Search Console lags two to three days, so the last few days are low.

In MochiStamp, Japanese pages have more impressions (833) than English (673) and Traditional Chinese (637), but not the most clicks: Traditional Chinese has 28 against Japanese's 19. Japanese pages average position 9.6, slightly behind English (8.9) and Traditional Chinese (8.2). In Gamma AI Slide, /ja/ pages average position 6.0, far ahead of the English pages' 27.4, but with only 56 impressions and 4 clicks, 2 of the clicks coming from legal pages like /ja/terms, so this is not evidence that Japanese product pages work.

Which Japanese pages have impressions?

Of MochiStamp's Japanese pages, the 8 with the most impressions are below (same window). Almost all target a specific question Japanese users search for: LINE sticker specs, how to write the description, transparent backgrounds, phrases.

Japanese page (MochiStamp, no prefix)ImpressionsClicksAverage position
/ideas/new-year146310.2
/guide/size9727.9
/guide/setsumeibun8146.4
/ideas/63010.3
/guide/haikei-touka5707.6
/phrases/teinei4319.4
/guide/serifu2517.0
/phrases/kimochi2509.4

These 8 pages total 537 impressions, 64% of the Japanese pages' 833. Gamma AI Slide's /ja/ has 8 pages with impressions, only 56 in total:

Japanese page (Gamma AI Slide, /ja/)ImpressionsClicksAverage position
/ja/edit-powerpoint-with-ai3625.4
/ja/terms623.8
/ja/ai-slide-japanese408.3
/ja/refund4010.3
/ja/203.0
/ja/pdf-to-slides209.0
/ja/pricing107.0
/ja/privacy1011.0

Both sites follow the same pattern: pages with impressions are guide or feature pages aimed at one clear question, and the home and pricing pages get few. Note the small sample: Gamma AI Slide's Japanese PowerPoint editing page alone accounts for 64% of all /ja/ impressions.

What do Japanese queries look like?

MochiStamp has 171 queries in total, 39 of them containing kana, with 135 impressions and 2 clicks. Below are the 9 with the most impressions (as Search Console shows them, not rewritten):

QueryImpressionsClicksAverage position
line スタンプ セリフ 一覧2208.3
lineスタンプ 説明文 例1426.5
lineスタンプ 説明文 コツ808.4
line スタンプ 作り方6090.2
スタンプ セリフ608.7
lineスタンプ タイトル 例507.0
lineスタンプ 言葉 一覧5010.6
lineスタンプ説明文 英語408.0
透過スタンプ 使い方409.3

Three things stand out in this table:

  • Queries are short combinations: LINE plus things you fill in when submitting (「セリフ」「説明文」「タイトル」), plus 「例」「一覧」「コツ」.
  • The same meaning is written in different ways, and spaces and letters are inconsistent (「lineスタンプ」 and 「line スタンプ」 count as separate queries), so the page should contain several of these forms, not target only one.
  • The query at position 90 is almost never seen by anyone; the impression was only counted, so it cannot be read as "having a ranking".

None of Gamma AI Slide's 85 queries contains kana, and its 56 Japanese-page impressions have no query text; we did not look into why. So the "how users search" section for Japanese comes from one site only, MochiStamp.

How should lang, hreflang and title be written on Japanese pages?

On 2026-11-12 we opened both sites' live Japanese pages with curl and listed how they are actually written. This is what they do, not a Google rule; the counts are characters when title and description are counted as strings.

ItemMochiStamp Japanese pagesGamma AI Slide /ja/ pages
html taglang is ja, dir is ltrlang is ja, dir is ltr
og:localeja_JPja_JP
canonicalPoints to itself (Japanese URL with no prefix)Points to itself (/ja/ URL)
Pages with several language versions: hreflangja, en, zh-Hant, zh-Hans, plus x-default pointing to the English pageen, ja, zh-Hans, zh-Hant, ru, plus x-default pointing to the English page
Pages with only a Japanese version: hreflangOnly ja and x-default (both pointing to itself)Only ja and x-default (both pointing to itself)
Title length (3 pages and 2 pages sampled)41, 43, 5149, 64
Description length101, 131, 181106, 106
Japanese URLs in the sitemap151, 15% of 1,02022, 14% of 156

For the principles of writing titles and descriptions see Do title and meta description matter for AI search?, and for how to write hreflang in full see How do you write hreflang on a multilingual site? How does AI search treat it?. We did not test how long a title can be before Google truncates it, so we give no length limit.

Where do visitors to Japanese pages come from?

GA4 (same window) by language prefix of the landing page: MochiStamp has 591 sessions in total, 258 of them (44%) landing on Japanese pages, 106 on English and 47 on Traditional Chinese. Gamma AI Slide has 319 sessions in total, only 14 landing on /ja/ pages and 238 on the English pages with no prefix.

By country (daily active users summed across days, not distinct users): MochiStamp has 103 from Japan and 135 from the United States; Gamma AI Slide has 9 from Japan. The dashboard cannot cross landing pages with sources, so we have no data on whether ChatGPT or Google's AI answers sent users to Japanese pages, and we do not claim any.

Building Japanese pages: what to do first?

  1. 1. Write about specific questions Japanese users search for

    Evidence: MochiStamp's most-shown Japanese pages target specific problems met when submitting stickers, such as specs, descriptions, transparent backgrounds and phrases, for example MochiStamp's Japanese sticker spec page and MochiStamp's Japanese description example page.

  2. 2. Write the page in Japanese itself and declare the language

    Set lang to ja, og:locale to ja_JP, and canonical to the page itself.

  3. 3. List only versions that exist in hreflang

    A page with only a Japanese version lists only ja and x-default; point to other versions only when they exist.

  4. 4. Put the forms users type into the title

    Evidence: the same phrase is written as LINE, line, ライン, with and without spaces, and the query table counts them separately.

  5. 5. Request indexing for new pages, then look at two to three weeks of data

    Search Console limits how many URLs you can submit each day, so submit the few pages you most want found first; see How do you get a new website indexed by Google? How many indexing requests can you send a day?.

What are the limits of these numbers?

  • Only two sites and four weeks of data; Gamma AI Slide's 8 Japanese pages total just 56 impressions.
  • The page table and the query table are separate; queries cannot be matched to a specific page.
  • The "other" bucket has no URL, and 1,440 of MochiStamp's impressions fall in it; the language table does not count them.
  • We did not check how many competing pages exist for these queries, so there is no conclusion on why positions are high or low.
  • We did not have a native Japanese speaker review page quality.

The same site's other languages are covered in Arabic and Persian pages in search and AI answers: what one small site saw and Traditional Chinese pages in search and AI answers: real data from two sites.

Where search results from Japan's two entry points, Yahoo! JAPAN and docomo, come from and whether to optimize for them separately is in Where do Yahoo! JAPAN and docomo search results come from? Do Japanese indie sites need to optimize for them separately?.

FAQ

Do Japanese pages do better than English on Google?

Not clearly, in MochiStamp: Japanese pages average position 9.6 with 19 clicks, English pages 8.9 and 8, Traditional Chinese pages 8.2 and 28. Two sites and four weeks of data, so do not copy it as is.

How do I write hreflang for Japanese pages?

Write ja. On these two sites' Japanese pages, those with several language versions list every version that exists plus x-default (pointing to the English page), and those with only Japanese list only ja and x-default.

Should Japanese pages use URLs with no language prefix?

MochiStamp's Japanese pages have no prefix and Gamma AI Slide's are under /ja/; both are being indexed. We have not compared which is better.

How long should a title be?

The Japanese pages we sampled on these two sites have titles of 41 to 64 characters. We did not test how long a title can be before it is truncated, so we give no limit.

Should LINE and line be written separately in Japanese queries?

In the query table, 「lineスタンプ」 and 「line スタンプ」 are counted as two separate queries. It is best for several forms to appear naturally on the page, but do not force them in just to match.

Are Japanese pages cited by ChatGPT?

No data can confirm it. The dashboard's ChatGPT sessions can only be attributed to the whole site, not to a page or language.

Why do Gamma AI Slide's Japanese pages have so few impressions?

This site's /ja/ section has only 22 URLs in the sitemap, and we did not check search volumes for similar terms in Japanese, so there is no conclusion on the cause.

Sources

kagentx

Do AI answers
mention you?

An evaluation asks ChatGPT, Perplexity and Claude your buyers' questions (with web search on) and records who they recommend, which pages they cite and whether you are among them. Basic $1, results usually in 2–3 minutes.