The Domains ChatGPT Cites in Japan: Why Reddit, PR TIMES, note and Rakuten Top the List, and How to Map Citation Sources for Your Own Market

Key Takeaways: ChatGPT's Citation Map Is Two Layers, Global Reference Sources Plus Local Publishing Platforms

The first question I get about ChatGPT optimization is always "where do I need to be listed to get cited?" I lined up Ahrefs' Japan data across three points in time, from May 2025 to June 2026, and cross-checked it against their industry breakdown. Japan is the market we track most closely, but the structure generalizes. Here's what emerged.

1. The top of the list mixes global reference sources with local publishing platforms — ChatGPT's top 10 cited domains for Japanese queries in June 2026 were Reddit, PR TIMES (press release wire), Ameblo (blog platform), English Wikipedia, note (long-form platform), Rakuten marketplace, Mynavi News, YouTube, Yahoo! Chiebukuro (Q&A), and ITmedia. Other AI engines put Japanese Wikipedia and note at the top consistently. Only ChatGPT blends in English-language sources
2. Rankings reshuffle within three months — Ameblo was first in April 2026 and third in June. Reddit went from eighth to first. YouTube and Yahoo! Chiebukuro entered the top 10 in June. Optimizing for "whoever is first right now" doesn't hold up
3. Every industry has a different staple — In Ahrefs' industry analysis (September to December 2025), ChatGPT's top cited domain was Reddit for IT, PR TIMES for food, and English Wikipedia for automotive. You have to draw the map for your own category
4. English Wikipedia outranking Japanese Wikipedia points to English sub-query expansion — Ahrefs suggests ChatGPT expands Japanese questions into English sub-queries internally. MJ Cachón's brand-prompt study found 27.3% of sub-queries for international brands contained English vocabulary
5. Only 32.3% of ChatGPT's top 1,000 cited URLs are within marketers' direct reach; the remaining roughly 67% are not — In Ahrefs' September 2025 analysis (English-language data), the main non-influenceable buckets were Wikipedia at 29.7%, homepages at 23.8% and app stores at 6.6%. Focus on the influenceable 32.3% (educational content, reviews, news, blogs), and treat the homepage share as "fix your own official site"

Below, I go through the timeline, the industry view, the language question and the influence question, and close with a method for mapping your own category.

---

1. Three Snapshots of ChatGPT's Citation Sources in Japan

Ahrefs publishes rankings of domains cited by AI search in Japan based on its Brand Radar data. Three snapshots are public for ChatGPT.

Numbers in parentheses are published citation counts. The periods and counting conditions differ between snapshots, so the counts aren't directly comparable. What matters is the cast and the movement.

Stable cast, unstable order

Reddit, English Wikipedia, Ameblo, PR TIMES and note appear across all three snapshots. The order, though, moves a lot. Reddit was first in 2025, out of the top five (eighth) in April, and back to first in June. Ameblo rose to first in April and dropped to third in June.

Honestly, this is where I stopped thinking in terms of "optimizing for ChatGPT's citation ranking." Build tactics around a metric whose leader changes every three months and the premise is gone by the time you publish. Cover the stable cast, then track who's actually cited for your category rather than the overall order.

Volatility itself is the subject of ChatGPT's Citation Sources Moved 86% in Four Days.

References: Ahrefs analysis of AI search citation sources in Japan (Japanese) - Manamina Top 10 domains cited by AI search, April 2026 update (Japanese) - Ahrefs Top Japanese domains cited by AI, June 2026 (Japanese) - Ahrefs

---

2. Every Industry Has a Different Staple

The overall ranking hides the industry differences. Ahrefs' industry analysis, published November 30, 2025, covered September 2 to December 1, 2025 across automotive, IT and food, for AI Overviews, AI Mode and ChatGPT. Here's the ChatGPT slice.

Here's what's interesting: three industries, three different leaders. IT is dominated by Reddit, where engineers argue in public. Food is led by press releases and marketplace product pages. Automotive is English Wikipedia plus two specialist car sites.

So neither "ChatGPT cites Reddit" nor "ChatGPT cites press releases" is accurate without naming the industry. The map for your category can only be drawn from queries in your category.

For contrast, AI Mode had YouTube first in all three industries, and AI Overviews had YouTube first in two of three. Google's engines and ChatGPT eat completely different diets.

References: 2025 AI search trend analysis: industry citation patterns in Ahrefs Brand Radar (Japanese) - Ahrefs

---

3. Why English Wikipedia Outranks Japanese Wikipedia

This is the most overlooked point in Japanese ChatGPT optimization, and it carries a lesson for any non-English market. The questions are in Japanese, yet English Wikipedia sits near the top of the citation list and Japanese Wikipedia doesn't appear at all. The pattern holds in June, in April, and in the 2025 data.

Ahrefs' June commentary suggests ChatGPT expands the original query into English before searching. When I cross-checked that against other observational data, it held up. MJ Cachón's August 2026 study of 723 brand-prompt runs through OpenAI's API found English vocabulary in 8.9% of all sub-queries and in 27.3% for international brands.

If ChatGPT decomposes a Japanese question into English sub-queries, a site that only exists in Japanese can't be found by them. Conversely, a site with English summaries or an English edition gains an additional path into the answer for a Japanese question.

What to do about it

• Add English summaries (even just headings and conclusions) to key pages and connect them to the local version with hreflang
• Write proper nouns in both the local script and English (company, product, technical terms)
• Check whether English Wikipedia has an entry for your brand or category. If not, don't wait for someone to write one; publish citable primary information (press releases, research) in English as well

The fan-out mechanics and the measured English mix are covered in What Happened to ChatGPT's Query Fan-Out.

References: Top Japanese domains cited by AI, June 2026 (Japanese) - Ahrefs A study of query fan-out in ChatGPT using brand searches - MJ Cachón

---

4. About 67% of Citations Are Out of Reach. Focus on the 32.3% That Isn't

Once you have the map, separate what you can actually reach. Ahrefs' October 28, 2025 analysis classified ChatGPT's top 1,000 cited URLs from September 2025 by page type (English-language data).

Ahrefs groups educational content, reviews, news and blogs (32.3%) as citations marketers can influence, and puts the group they can't directly influence at roughly two thirds (about 67%). The rows above are Ahrefs' categories transcribed as published; the buckets overlap, so they sum to more than 100% and the 67% is not a simple complement of the rows. Read this as "about a third is influenceable" rather than adding up individual rows.

I'd add one qualification. You can't influence other companies' homepages, but you can fix your own. With 16.6% of ChatGPT's sub-queries for brand prompts searching the brand's own domain by name, making your homepage and key pages clearly state who you are and what you offer is how you claim your slice of that 23.8%.

Translated to the Japanese market, the influenceable territory reads like this.

• Educational content — Your own blog, explainer articles on note, contributed pieces on ITmedia or Mynavi News
• Reviews — Rakuten and Amazon reviews, comparison sites, answers on Yahoo! Chiebukuro
• News — Press releases on PR TIMES, trade media coverage
• Homepage — The structure of your official site

For other markets, substitute the local equivalents: your press release wire, your dominant long-form platform, your Q&A site, your marketplaces.

References: ChatGPT's most cited pages - Ahrefs

---

5. How ChatGPT Differs from Other AI Engines: Only ChatGPT Mixes in English Sources

If you assume the same tactics work for ChatGPT, Google AI Overviews and Perplexity, you'll miss. Ahrefs' April 2026 data lists each engine's top three cited domains for Japanese queries.

Google's engines and Perplexity center on YouTube and Japanese Wikipedia. Copilot leans on Japanese Wikipedia and note. Only ChatGPT puts a press release wire and English Wikipedia at the top. That difference presumably comes from differences in the indexes and retrieval pipelines behind each engine. For ChatGPT's three-layer index, see Inside ChatGPT Search.

The practical conclusion is to log citation sources per engine. Being cited in ChatGPT but not in AI Overviews, or the reverse, is common. LinkSurge's GEO monitoring records, for each watch query, whether your brand is mentioned and which URLs are cited, separately for ChatGPT Search, Gemini, Google AI Overviews, Claude and Perplexity.

References: Top 10 domains cited by AI search, April 2026 update (Japanese) - Ahrefs

---

6. How to Map Citation Sources for Your Own Category

Let's turn this into a procedure. Treat the overall ranking as background and draw your own map.

1. Design 20 to 30 questions — Shortlist-style queries first: "best [category]," "[A] vs [B]," "how to choose [category] for [use]"
2. Run each in ChatGPT at least three times and log the cited URLs — Answers vary run to run, so one pass is skewed
3. Group cited URLs by domain into five buckets — Reference sources (Wikipedia etc.), press releases, UGC and blogs (note, Ameblo, Q&A sites, Reddit), marketplaces and reviews, trade media
4. Prioritize domains where competitors are cited and you aren't — Restrict to the influenceable buckets: educational content, reviews, news, your own site
5. Re-measure monthly with the same questions — The order changes within three months, so keep the map current

Steps 2 and 5 are tedious by hand, which is why we register watch queries and let the logging run automatically. Per-engine records also reveal splits like "PR TIMES worked in ChatGPT, note worked in AI Overviews."

The full strategy is in How to Get Cited by ChatGPT: The Complete Guide. How to use press release wires, long-form platforms and Q&A sites specifically is in Using Press Releases, note and Q&A Sites for ChatGPT Citations.

---

Frequently Asked Questions

Which domains does ChatGPT cite most for Japanese queries?

In Ahrefs' June 2026 data, the top five were Reddit, PR TIMES, Ameblo, English Wikipedia and note. In April 2026, Ameblo was first and Reddit was eighth, so the order changes within months. The stable cast across snapshots is Reddit, English Wikipedia, Ameblo, PR TIMES and note.

Do ChatGPT's citation sources vary by industry?

Yes. In Ahrefs' September to December 2025 industry analysis for Japan, ChatGPT's top cited domain was Reddit for IT (5,439 responses), PR TIMES for food (2,445) and English Wikipedia for automotive (2,289). You need to check which domains are actually cited for questions in your own category rather than relying on the overall ranking.

Why does ChatGPT cite English Wikipedia for Japanese questions?

Ahrefs suggests ChatGPT expands Japanese questions into English sub-queries internally. MJ Cachón's August 2026 study found English vocabulary in 27.3% of sub-queries for international brands. A site that exists only in the local language can't be found by those sub-queries, so English summaries of key pages and English spellings of proper nouns are the countermeasure.

How much of ChatGPT's citations can a brand actually influence?

In Ahrefs' September 2025 English-language data, 32.3% of ChatGPT's top 1,000 cited URLs (educational content, reviews, news, blogs) were directly influenceable, and Ahrefs classifies the remaining roughly 67% (mainly Wikipedia 29.7%, homepages 23.8% and app stores 6.6%) as not. Since your own homepage is part of the 23.8% homepage share, improving your official site's structure is a tactic you can act on.

Are ChatGPT's citation sources the same as Google AI Overviews'?

No. In Ahrefs' April 2026 data, ChatGPT's top sources were Ameblo, PR TIMES and English Wikipedia, while Google AI Overviews cited YouTube, Japanese Wikipedia and note, and AI Mode cited YouTube, Google and Instagram. Log citation sources per engine and plan tactics per engine.

---

Conclusion: Don't Chase the Overall Ranking. Draw Your Category's Map

ChatGPT's citation sources in Japan form a two-layer structure of global reference sources and local publishing platforms. The cast is stable; the order reshuffles within three months. Each industry has a different staple, and only about a third of the top 1,000 citations sit within marketers' direct reach.

That's why the work isn't optimizing for the overall ranking. It's logging who gets cited for your category's questions, concentrating on the influenceable territory (educational content, reviews, news, your own site) and on English landing spots, and refreshing the map monthly.

LinkSurge's GEO monitoring records cited URLs per watch query across five engines including ChatGPT Search, with competitor comparison. Use it for the first map and the monthly refresh.