- 01
Across 5 AIs × 5 scenarios × 5 runs — 125 responses and 625 recommendation slots — 621 recommendations could be extracted as property names, covering 114 properties. As a reference comparison, this corresponds to roughly 25.1% of the 454 ryokan and hotels in the town of Hakone.
- 02
The top 5 properties accounted for 31.1% of the 621 valid recommendations, showing a tendency for AI recommendations to concentrate on a limited set of properties. Fujiya Hotel was the most recommended (51 mentions), followed by Hakone Kowakien Ten-yu (49) and Hakone Ashinoko Hanaori (34). Gora Kadan received the most #1 recommendations (19).
- 03
Recommendations shifted clearly by travel scenario: Hakone Hotel Kowakien for families with children, Gora Kadan for inbound travelers, and Fujiya Hotel for adult family trips. Perplexity returned 4 non-property answers (booking sites, curated lists, and similar) instead of specific properties.
- Survey name
- AI Hotel Recommendation Survey (Hakone Edition)
- Subjects
- Hotel and ryokan recommendations for Hakone by major generative AI services
- AI services
- ChatGPT, Claude, Gemini, Perplexity, Microsoft Copilot
- Survey period
- July 1–2, 2026
- Runs
- 5 AIs × 5 scenarios × 5 runs, 125 responses (625 recommendation slots)
- Data captured
- Recommendation rank, property name, reason, best-fit traveler, suggested booking channel, URL
- Method
- Normalized name variants, then aggregated recommendation counts, #1 counts, ranking points, and trends by scenario and by AI service
- Conducted by
- Stellagent Inc.
Summary
Stellagent Inc. surveyed which hotels and ryokan the major generative AI services recommend when asked about accommodation for a trip to Hakone, and in which travel scenarios each property tends to appear.
We set up five scenarios — a couple's trip, a family trip with children, an inbound (first-visit) trip, a trip with a group of young friends, and an adult family trip with parents — and asked the web UIs of ChatGPT, Claude, Gemini, Perplexity, and Microsoft Copilot the same questions five times each. The survey maps how accommodation providers appear on AI and which travelers they get recommended to, at a time when AI is becoming the entry point for choosing a hotel.
| Metric | Result |
|---|---|
| AI services | 5 |
| Scenarios | 5 |
| AI responses | 125 |
| Recommendation slots | 625 |
| Recommendations extracted as property names | 621 |
| Non-property answers | 4 |
| Distinct properties recommended | 114 |
| Ryokan and hotels in the town of Hakone | 454 |
| Reference share of properties appearing in AI answers | approx. 25.1% |
| Recommendations for the top 5 properties | 193 |
| Share held by the top 5 properties | 31.1% |
While recommendations were spread across 114 properties, the top 5 properties accounted for 31.1% of the 621 valid recommendations, showing a tendency for AI recommendations to concentrate on a limited set of properties. Fujiya Hotel was the most recommended with 51 mentions, followed by Hakone Kowakien Ten-yu with 49 and Hakone Ashinoko Hanaori with 34. Gora Kadan received the most #1 recommendations, at 19.

The chart above may be reproduced as-is in media coverage, articles, and other materials, provided that you credit "Stellagent Inc." as the source and include a link to this page.
Recommendations also shifted clearly by travel scenario. Hakone Hotel Kowakien stood out for families with children, Gora Kadan for inbound travelers, and Fujiya Hotel for adult family trips — even within Hakone, the candidates AI raises change depending on who you travel with.
Background
Travelers no longer only compare hotels by entering conditions on booking sites. They increasingly ask AI questions like "Which ryokan in Hakone is best for couples?" or "Which hotel should we pick for a Hakone trip with elementary-school kids?"
The properties that appear in AI answers become the traveler's first shortlist. For accommodation providers, this means that rankings on search engines and OTAs are no longer the only thing that matters. Whether AI recommends your property, how it describes which travelers you suit, and whether it connects users to your official booking flow are now in question as well.
We conducted this survey of the Hakone area to clarify these emerging issues for hotel marketing in the age of AI agents.
Method
For each of the five travel scenarios, we entered the same prompt into each AI service's web UI five times, obtaining 125 responses and 625 recommendation slots in total. From each response we recorded the recommendation rank, property name, reason, best-fit traveler, suggested booking channel, and URL. For aggregation, we normalized name variants and tallied recommendation counts, #1 recommendation counts, ranking points, and trends by scenario and by AI service.
Controlling for account context
To prevent account-specific context from influencing answers, we used Temporary mode, temporary chats, or fresh chats in each AI service's web UI. Each prompt also explicitly instructed the AI not to reference past conversations or memories about the user.
Notes on the survey
This is an exploratory survey based on AI answers as of the survey dates. AI answers can vary by date and time, model, search settings, conversation history, region, account settings, and other factors. This survey does not evaluate the relative quality of any hotel or ryokan.
Scenarios
| ID | Scenario | Conditions |
|---|---|---|
| S1 | Couple's trip | 1 night from Tokyo, two adults (married couple or couple) |
| S2 | Family trip with children | 1 night from Tokyo, two adults and two elementary-school children |
| S3 | Inbound trip | First visit to Japan, one night in Hakone with a partner, English prompt |
| S4 | Young friends' trip | 1 night from Tokyo, four university students or friends in their 20s |
| S5 | Adult family trip | Four adults — parents and their adult children — one night in Hakone |
As a rule, the prompts did not specify preferences such as hot springs, meals, price, access, quietness, or photogenic appeal. This was to observe which accommodation values the AI chooses to emphasize on its own.
Prompts
The following prompts were entered into each AI service per scenario. Each prompt opens by instructing the AI not to reference past conversations or memories, to avoid account-specific context. S1, S2, S4, and S5 were asked in Japanese (shown here in English translation); S3 was asked in English as written.
S1 Couple's trip
For this question, please do not reference past conversations or any memories about me; use only the information in this message.
We are taking a 1-night, 2-day trip from Tokyo to Hakone. We are two adults, a married couple or couple. We have not decided which hotel or ryokan to stay at in Hakone.
Please recommend five options. For each, provide 1) rank, 2) hotel or ryokan name, 3) reason, 4) best-fit traveler, and 5) which official website or booking site to check, in table format. Include URLs if you know them; if not, write "URL unknown."
S2 Family trip with children
For this question, please do not reference past conversations or any memories about me; use only the information in this message.
We are taking a 1-night, 2-day family trip from Tokyo to Hakone with two adults and two elementary-school children. We have not decided which hotel or ryokan to stay at in Hakone.
Please recommend five options. For each, provide 1) rank, 2) hotel or ryokan name, 3) reason, 4) best-fit traveler, and 5) which official website or booking site to check, in table format. Include URLs if you know them; if not, write "URL unknown."
S3 Inbound trip
Please answer without referring to any past conversations or memories about me. Use only the information in this message.
I am visiting Japan for the first time and want to stay in Hakone for one night with my partner. We have not decided which hotel or ryokan to book yet. Please recommend five hotels or ryokan in Hakone.
For each recommendation, provide 1) rank, 2) hotel or ryokan name, 3) reason, 4) best-fit traveler, and 5) the official website or booking site to check. Include URLs if you know them. If you do not know the URL, write "URL unknown."
S4 Young friends' trip
For this question, please do not reference past conversations or any memories about me; use only the information in this message.
We are taking a 1-night, 2-day trip from Tokyo to Hakone as a group of four university students or friends in their 20s. We have not decided which hotel or ryokan to stay at in Hakone.
Please recommend five options. For each, provide 1) rank, 2) hotel or ryokan name, 3) reason, 4) best-fit traveler, and 5) which official website or booking site to check, in table format. Include URLs if you know them; if not, write "URL unknown."
S5 Adult family trip
For this question, please do not reference past conversations or any memories about me; use only the information in this message.
We are taking a 1-night trip to Hakone as a family of four adults — parents and their adult children. We have not decided which hotel or ryokan to stay at in Hakone.
Please recommend five options. For each, provide 1) rank, 2) hotel or ryokan name, 3) reason, 4) best-fit traveler, and 5) which official website or booking site to check, in table format. Include URLs if you know them; if not, write "URL unknown."
Recommendation Ranking
| Rank | Property | Recommendations | #1 recommendations | Ranking points | Main scenarios |
|---|---|---|---|---|---|
| 1 | Fujiya Hotel | 51 | 13 | 174 | Adult family, family with children |
| 2 | Hakone Kowakien Ten-yu | 49 | 10 | 161 | Family with children, adult family |
| 3 | Hakone Ashinoko Hanaori | 34 | 9 | 104 | Couples, family with children |
| 4 | Tenseien | 30 | 1 | 77 | Family with children, young friends |
| 5 | Hakone Yutowa | 29 | 13 | 103 | Young friends, couples |
| 6 | Hakone Ginyu | 25 | 8 | 94 | Inbound, couples |
| 7 | Hakone Hotel Kowakien | 24 | 16 | 107 | Family with children, young friends |
| 8 | Tsuki no Yado Sara | 24 | 1 | 58 | Couples, family with children |
| 9 | Gora Kadan | 23 | 19 | 107 | Inbound, young friends |
| 10 | Hakone Suishoen | 23 | 0 | 65 | Adult family, family with children |
| 11 | Hotel Okada | 20 | 2 | 60 | Family with children, adult family |
| 12 | Toki no Yu Setsugetsuka | 17 | 0 | 36 | Adult family, family with children |
| 13 | Yama no Chaya | 14 | 0 | 32 | Inbound, couples |
| 14 | Fufu Hakone | 14 | 2 | 48 | Couples, inbound |
| 15 | Hakone Hotel | 14 | 1 | 45 | Young friends, family with children |
*Ranking points were tallied as 5 points for #1, 4 for #2, 3 for #3, 2 for #4, and 1 for #5.
According to materials published by the town of Hakone, there were 454 ryokan and hotels in the town as of fiscal 2024. The 114 properties that appeared in AI answers in this survey correspond to roughly 25.1% of that figure as a reference comparison. Note that AI answers may include properties outside the town of Hakone or properties whose location requires verification, so this ratio is a reference value prior to location vetting.
*The number of ryokan and hotels in the town is based on p.3 of "Materials 1: Current State of Tourism and Fiscal Outlook," 2nd Hakone Tourism Community Development Funding Review Meeting, town of Hakone (held August 6, 2025; PDF, confirmed July 8, 2026).
Trends by Scenario
S1 Couple's trip
| Property | Recommendations |
|---|---|
| Tsuki no Yado Sara | 11 |
| Hakone Ashinoko Hanaori | 10 |
| Hakone Kowakien Ten-yu | 8 |
| Hakone Yutowa | 8 |
| Fufu Hakone | 7 |
For couples, recommendations leaned on contexts such as open-air baths in guest rooms, modern stays, views, and anniversary use, along with relatively new properties and easy-to-stay resort-style facilities.
S2 Family trip with children
| Property | Recommendations |
|---|---|
| Hakone Hotel Kowakien | 16 |
| Hakone Kowakien Ten-yu | 15 |
| Tenseien | 14 |
| Hotel Okada | 14 |
| Fujiya Hotel | 12 |
For families with children, the AIs emphasized play elements such as the Yunessun spa resort, access from stations, kid-friendliness, buffets, and the reassurance of large ryokan.
S3 Inbound trip
| Property | Recommendations |
|---|---|
| Gora Kadan | 15 |
| Hakone Ginyu | 14 |
| Yama no Chaya | 10 |
| Hakone Kowakien Ten-yu | 9 |
| Hyatt Regency Hakone | 8 |
For inbound travelers, properties offering a traditional ryokan experience, expected English support, historical character, and international recognition tended to be recommended.
S4 Young friends' trip
| Property | Recommendations |
|---|---|
| Hakone Yutowa | 13 |
| Tenseien | 9 |
| Fujiya Hotel | 7 |
| Hakone Ashinoko Hanaori | 7 |
| Hakone Hotel | 6 |
For groups of young friends, the AIs valued price, access, large communal baths, buffets, on-site facilities, and rooms that groups can easily share.
S5 Adult family trip
| Property | Recommendations |
|---|---|
| Fujiya Hotel | 21 |
| Hakone Kowakien Ten-yu | 12 |
| Toki no Yu Setsugetsuka | 9 |
| Hakone Suishoen | 7 |
| Hakone Yutowa | 6 |
For adult family trips, support went to access and room size, plus historic properties, easy access to hot springs, and dining formats that work even when preferences differ across generations.
Differences by AI Service
| AI service | Valid recommendations | Distinct properties | Most-recommended properties |
|---|---|---|---|
| ChatGPT | 125 | 32 | Hakone Yutowa, Hakone Suishoen, Fujiya Hotel |
| Claude | 125 | 35 | Fujiya Hotel, Hakone Yutowa, Hakone Kowakien Ten-yu |
| Gemini | 125 | 32 | Hakone Kowakien Ten-yu, Fujiya Hotel, Hakone Ashinoko Hanaori |
| Perplexity | 121 | 51 | Tenseien, Tsuki no Yado Sara, Fujiya Hotel |
| Copilot | 125 | 31 | Hakone Ginyu, Hakone Ashinoko Hanaori, Hakone Suishoen |

The chart above may be reproduced as-is in media coverage, articles, and other materials, provided that you credit "Stellagent Inc." as the source and include a link to this page.
The properties recommended differed by AI service. Perplexity recommended the most distinct properties at 51, spreading its recommendations relatively widely. Copilot, ChatGPT, and Gemini each stayed in the low 30s, concentrating recommendations on a more limited set of properties.
Non-Property Answers and Extraction Notes
| Category | Count | Treatment |
|---|---|---|
| Recommendations extracted as property names | 621 | Included in the property ranking |
| Non-property answers | 4 | Excluded from the property ranking |
| Total recommendation slots | 625 | All slots obtained from AI answers |
Non-property answers were returned not as specific hotel or ryokan names but as booking sites, groups of properties, curated feature pages, and the like. This confirms that even when users ask for property names, not every slot in an AI answer necessarily returns a specific property. All 4 such answers in this survey came from Perplexity.
We also recorded the booking sites and URLs included in answers, but the booking-channel classification for the 125-response dataset is outside the scope of the finalized tallies at the time of publication. This page focuses on the concentration of property recommendations and trends by scenario.
Implications for Hotel and Ryokan Operators
As AI agents come to assist travelers with hotel selection, accommodation providers will need to respond in the following ways.
1. Organize property information so AI can understand it
It is important to clearly organize, on your official website, information such as rooms, hot springs, dining, access, suitability for children, group friendliness, and fit for multi-generation trips.
2. Check how you appear in each scenario
The same property looks different on AI depending on the scenario — couples, families with children, inbound travelers, young friends, or adult family trips. You need to know in which scenarios you get recommended and in which you rarely appear.
3. Make your official booking flow easy to reach from AI
Even when an AI answer presents your official site, it is not guaranteed that users can smoothly proceed to availability search, plan comparison, membership sign-up, and booking completion. Designing the path from AI-mediated travel consultation to official booking as a single flow is essential.
4. Reduce information that is easily misunderstood
Features limited to specific plans, seasons, or room types may be overgeneralized in AI answers. Clearly stating conditions on your official site reduces gaps with traveler expectations.
Notes and Limitations
- This is an exploratory survey of 5 AIs × 5 scenarios × 5 runs and is not intended for statistical generalization.
- AI answers can vary by date and time, model, search settings, conversation history, region, account settings, and other factors. This report is based on AI answers as of the survey period (July 1–2, 2026).
- Prices, availability, and plan contents fluctuate and were not evaluated.
- The comparison with the number of ryokan and hotels in the town of Hakone is a reference value prior to location vetting, as AI answers may include properties outside the town or properties whose location requires verification.
- This survey does not evaluate the relative quality of any hotel or ryokan.
- Product and service names mentioned in this report are trademarks or registered trademarks of their respective owners.
Revision History
- July 8, 2026: First edition published
You are free to cite or reproduce the findings of this report in articles, media coverage, and other materials, provided that you credit “Stellagent Inc.” as the source and include a link to this page.
Free AI-Agent Readiness Assessment
As AI search and LLMs become the entry point to booking and purchasing, we assess — for free — whether your official site is correctly read and chosen by AI. You'll receive a summary of your current AI-readability and the improvements to prioritize so you get chosen via AI. If you're interested, apply below.
