GMO NIKKO’s GMO AI Search Lab asked 1,667 Japanese consumers aged 15 to 59 who use Google AI Overviews where they go first when researching a product or service. Google Search still led at 79.5%, but 31.3% named generative AI, about the same as social media (35.1%) and EC sites (31.5%). That changes when and how your tsuyu content strategy should be written. The shopper looking for heya-boshi fixes next June may never scroll a results page at all. They’ll read an AI answer and decide.
A tsuyu content strategy for AI search means publishing structured, question-first pages about rainy-season problems (indoor laundry, mold, humidity) weeks before the Japan Meteorological Agency announces tsuyu-iri. Each page should give a direct answer AI Overviews can quote, cover kanji, kana and casual spellings, and be updated visibly during the season.
Why does tsuyu content need a different AI search strategy?
Tsuyu content needs its own strategy because, once the humidity arrives, people stop looking up products and start asking questions about their problem. Tsuyu (梅雨) is the early summer rainy season that covers most of Japan from late May or early June until mid-July, driven by a stalled weather front, grey skies and heavy humidity.
The effect on spending is measurable. A Bank of Japan review by Kunihiro Akutsu and Yasutaka Koike built a “Macro Weather Index” and found that more rain lowers consumer spending, with effects that vary by product category. Spending doesn’t just fall when it rains, though. It moves. Searches shift from “dehumidifier price” to messy, worried, very human questions like “why does my laundry still smell after washing” (部屋干し 臭い) or “my room feels damp no matter what I do.”
Those long, conversational questions are exactly what AI Overviews and AI Mode are built to answer. Google added Japanese to AI Mode in September 2025, one of the first five non-English languages it supported. This will be the second tsuyu season where Japanese shoppers can ask Google a complex rainy-season question and get one combined answer instead of ten blue links.
What does the GMO NIKKO data actually say about AI-first shoppers?
The data says that once a shopper starts with AI, they often never leave it. In the GMO NIKKO survey, published October 16, 2026, 87.5% of the people who use generative AI first said they had finished their research inside AI alone. 81.0% had bought a product or service they researched that way. Overall, 70.3% said they trust AI answers either “very much” or “somewhat.”
One detail I love: the survey ran from May 29 to June 9, 2026, which overlaps with the 2026 Kanto-Koshin tsuyu-iri (around June 7). The respondents were answering about AI-first shopping in the exact window we’re talking about.
When should tsuyu content be published to get indexed before tsuyu-iri?
Publish and get indexed in early to mid May, then refresh in late May. Tsuyu-iri (梅雨入り) is the Japan Meteorological Agency’s announcement that the rainy season has started, and it triggers an immediate shift in what people search for.
According to the JMA’s rainy season data, the normal tsuyu-iri for Kanto-Koshin (which includes Tokyo) is around June 7 and the normal end is around July 19. Those “normal” dates are a 30-year average for 1991 to 2020. In 2026 the season matched them almost exactly: it started around June 7 and ended around July 20.
Here’s the part almost nobody writing about seasonal SEO mentions. The JMA’s announcements are preliminary. In June 2025, Kanto-Koshin was announced as starting around June 10. The JMA’s final figures now list the 2025 start as around May 22, almost three weeks earlier. The weather didn’t wait for the press release, and neither did the damp towels.
So my honest view: don’t tie your publishing date to the announcement. Tie it to the weather, and the weather can arrive in late May.
| Window (Kanto-Koshin) | What to do | Why |
|---|---|---|
| Early to mid May | Publish core tsuyu pages | Gives crawlers and AI systems time to index and pull from them |
| Late May | Refresh stats, add an “Updated” date | The season can arrive as early as May 22 (as in 2025) |
| Day JMA announces tsuyu-iri | Update the opening answer and FAQ for your region | Freshness signal right when demand jumps |
| Late June | Add mid-season Q&As from real customer questions | AI answers favor current, specific content |
| Around July 19 | Pivot to summer heat and post-tsuyu content | Demand moves to cooling and heat stress |
What content format gets a brand cited for heya-boshi and mold prevention queries?
The format that gets cited answers the seasonal question in the first sentence, then adds detail, spelling variants and real named products or services. Heya-boshi (部屋干し) is the Japanese habit of drying laundry indoors on rainy or humid days. It drives seasonal demand for special detergents, odor removers, dehumidifiers and clothes dryers.
Most competing coverage stops at tourist advice like packing a raincoat or buying a convenience store umbrella, and never explains how an AI answer engine decides what to cite. GEO (Generative Engine Optimization) means shaping your content and structured data so it shows up inside AI-generated answers. A tsuyu page built for GEO should include:
- A one-sentence direct answer to the seasonal question as the first line of each section
- FAQ follow-ups that cover kanji, hiragana and katakana variants (部屋干し / へやぼし, カビ / かび / 黴)
- Named products, service tiers or local businesses instead of vague category words
- A comparison table where it helps, since AI engines extract structured blocks more reliably than long prose
- A visible publish date and a visible update date before tsuyu-iri
How does text search intent differ from AI search intent during tsuyu?
Text search still sends people to listing pages. AI search sends them to explanations. This is how I see the split playing out for the three biggest tsuyu query types:
| Query type | Typical Google text search result | Typical AI search behavior |
|---|---|---|
| Heya-boshi odor | Product listing pages, coupon sites | Combined explanation of the cause plus a product comparison |
| Mold prevention | Retailer category pages | Step-by-step answer: cause, then solution |
| Dehumidifier purchase | Price comparison shopping results | Feature comparison across named models |
Tsuyu is the only season in Japan when your towels develop a personality. Unfortunately, it’s the personality of a gym bag left in a car. If your content takes three paragraphs to explain why that happens, the AI answer will cite whoever explained it in one sentence. Nobody reading 部屋干し 臭い at 11 PM wants your brand story first.
How do Google, Yahoo! JAPAN and Bing split tsuyu search traffic?
On phones, where most rainy-day questions get typed, Google dominates and Yahoo! JAPAN is a strong second. On desktop, Bing is now almost tied with Google. StatCounter’s mobile data for August 2026 shows Google at 82.13%, Yahoo! at 14.17% and Bing at 0.91%. Yahoo! JAPAN’s results have long been powered by Google’s search technology, so the two share an index, but the audiences behave differently. Yahoo!’s portal-style users scan listings. Google’s AI Overviews users read a combined answer.
On desktop in Japan, StatCounter recorded Google at 48.41% and Bing at 46.81% in June 2026. That matters for high-ticket research done at a desk, and it matters for ChatGPT too, because ChatGPT Search draws on Bing’s index. I explain how each engine finds and cites sources in ChatGPT vs Google AI Overviews vs Perplexity: Where Should an SME Show Up First?. If your tsuyu plan ignores Bing Webmaster Tools, you’re skipping a door that a lot of desktop users walk through.
So write one page that does both jobs: scannable headings and tables for listing-style readers, and clear, quotable answers for AI summaries.
How can AI agents personalize tsuyu campaigns across LINE and web channels?
AI agents can send humidity-related messages at the right moment and update on-page content as each region’s tsuyu-iri is confirmed. A retail brand can run a chatbot that greets its LINE subscribers with heya-boshi tips the week the JMA announces onset in their region, while the same system updates the FAQ section on the website.
This is how Ongito’s LINE AI chatbot works for Japanese businesses. It runs on your LINE Official Account, where customers already read messages every day, instead of in a website widget nobody opens. Speed is what decides seasonal campaigns. When Ongito built a WhatsApp inquiry agent for my own photography business, Soredenacho, our response time dropped from over 15 hours to under 30 seconds. A tsuyu rush is the same problem with more umbrellas. Demand jumps overnight, and whoever answers first wins the sale.
What are the most common mistakes brands make with tsuyu AI search content?
The most common mistake is publishing one generic seasonal post in late May and never touching it again. By the time Google’s AI Overviews or Yahoo! JAPAN crawl it again mid-season, it’s out of date.
The second mistake is ignoring kana and casual spellings. Mold is usually written カビ in katakana, but people also type かび, and product packaging sometimes uses 黴. A page that only matches one spelling misses natural-language AI questions phrased the other ways.
The third mistake, and one that competing guides consistently miss, is treating Google and Yahoo! JAPAN as one audience while forgetting Bing entirely. The mobile and desktop splits are different markets and call for different structural choices.
The fourth mistake is thinking schema alone will get you cited. FAQ markup helps machines read your page, but if the answer underneath is padded marketing copy, there’s nothing worth quoting.
What does purchase intent look like for electronics versus FMCG during June humidity?
Purchase intent splits sharply between high-ticket electronics and everyday FMCG (fast-moving consumer goods) during the humid tsuyu stretch. Electronics research starts earlier, and FMCG purchases happen close to the moment of discomfort.
Someone comparing dehumidifiers or clothes dryers is likely to research over several sessions and several AI-assisted comparisons before buying. That’s why electronics content should go live in early May, with comparison tables and full specs. Someone restocking heya-boshi detergent or odor spray usually searches and buys the same day. FMCG content should put fast, direct answers first, so AI engines can quote them at the exact moment of need.
Frequently Asked Questions about Tsuyu Content Strategy for AI Search
Question: When does tsuyu start in Tokyo?
Answer: According to the Japan Meteorological Agency, the normal tsuyu-iri for Kanto-Koshin, which includes Tokyo, is around June 7, and the season normally ends around July 19. These are averages for 1991 to 2020. Actual dates vary, and in 2025 the confirmed start was around May 22.
Questions: When should brands publish rainy season content in Japan?
Answer: Brands should publish core tsuyu content in early to mid May, and refresh it in late May with an updated date. This gives Google, Bing and AI systems time to index the pages before demand peaks, and covers years when the season starts early.
Question: What is heya-boshi and why does it matter for marketers?
Answer: Heya-boshi (部屋干し) is the Japanese practice of drying laundry indoors during rainy or humid weather. It creates seasonal demand for special detergents, odor removers and dehumidifiers. Many of the questions behind that demand are long and conversational, which makes them ideal targets for AI Overviews.
Question: Do Google AI Overviews and Yahoo! JAPAN need different content?
Answer: They share Google’s search index, but their users behave differently, so one page should serve both. Use scannable headings and tables for Yahoo! JAPAN’s listing-style readers, and clear, quotable answers in the first sentence of each section for AI Overviews.
Question: Is FAQ schema enough to get tsuyu content cited by AI?
Answer: No. FAQ schema helps machines read your page, but AI systems cite answers that are direct, specific and current. You need answer-first writing, named sources, real data and visible updates during the season, with schema supporting that work rather than replacing it.
Conclusion: making a tsuyu content strategy work for AI search, not just rankings
If you treat tsuyu as a box to tick on the content calendar, you miss what GMO NIKKO’s research shows: a real share of Japanese shoppers now start, and often finish, their research inside generative AI. A tsuyu content strategy for AI search works when it’s published before the Japan Meteorological Agency’s normal start window, written for Google AI Overviews, Yahoo! JAPAN and Bing at the same time, and answers each question in its first sentence instead of hiding it under marketing copy.
If you’re reading this in November, you have about six months. That’s plenty of time if you start now, and not much if you start in May. If you want to know whether AI engines can find and cite your brand today, request a free audit or contact Ongito to work together on your AI and agentic visibility before next season. The rain will arrive on its own schedule. Whether your content is already indexed and waiting is the only part you control.