GEO Opportunities for Korean OTAs: A Research Analysis of ChatGPT Search Behavior

Key Takeaways

• ChatGPT also searched in Korean for Korean-language questions. Korean search queries appeared in 106 of the 110 prompts analyzed (96.4%).
• There are top-10 spots that Korean OTAs can compete for. Of the 1,060 top 10 Google results for Korean search queries, 344 (32.5%) came from sites with a Domain Rating (DR) of 68 or lower.
• Ranking alone did not lead to citations. Pages focused on booking interfaces were rarely cited, even when they ranked near the top. A test page that put its recommendation upfront was cited in a ChatGPT answer just 2 days after launch, despite being on a new domain.

To find out what Korean OTAs such as Yanolja, MyRealTrip, and Yeogi Eottae need to be cited in ChatGPT accommodation recommendations, Anymorph entered 115 Korean-language travel prompts into ChatGPT in December 2025 and collected the web search queries ChatGPT ran. We then compared Google’s top results for the Korean queries with each site’s Ahrefs Domain Rating (DR).

Research findings at a glance

Condition	Finding	What it means
Does ChatGPT search in Korean?	Korean search queries appeared in 106 of the 110 prompts analyzed (96.4%)	Korean-language pages can be candidates in search results
Can Korean OTAs make the top 10?	344 of the 1,060 top 10 results (32.5%) had a DR of 68 or lower	Korean OTAs with DRs of 55–71 have room to compete
Is there content for AI to cite?	OTA pages that ranked near the top mainly featured booking interfaces	Pages with recommendations and explanations are needed

Research background

How does ChatGPT create answers using web search?

When ChatGPT uses web search, it turns a question into several search queries, reads the pages it finds, writes an answer, and adds citations.

1. Prompt submitted. The user asks a question in Korean.
2. Search queries generated. ChatGPT turns the question into several search queries. In this study, 115 prompts generated 346 queries, an average of about 3 per prompt.
3. Search results collected. Results pages are gathered for each query.
4. Sources selected. Pages are selected for use in the answer.
5. Answer generated. The answer is shown with links to its sources.

For this study, we searched Google using the Korean queries generated by ChatGPT and compared the top 10 results. The search results ChatGPT actually read may differ from Google’s rankings.

What is GEO?

GEO (Generative Engine Optimization) is the practice of creating content so that brands are mentioned and cited in answers generated by AI.

Category	SEO	GEO
Goal	Improve search-result rankings	Get brands mentioned and cited in AI answers
Key metrics	Rankings, click-through rate	Mention rate, citation frequency
Result	A link in a list of search results	AI directly recommends a brand in its answer

Why does this matter to Korean OTAs?

Metric	Figure	Source
Consumers already using AI search	About 50%	McKinsey, 2025
Value of visits from AI search (by conversion rate, compared with regular organic search)	4.4 times	Semrush, 2025
Share of Google searches with AI Overviews (March 2025)	18%	Pew Research Center
Visits that clicked regular results when an AI Overview was present (15% when there was no overview)	8%	Pew Research Center

Research methodology

Test design

We created 115 Korean-language travel prompts. Questions about accommodations in Japan were the most common, with 65; there were 20 about Southeast Asia, 15 about Korea, and 15 about Europe and the Americas (including Guam and Saipan). The full list is in the appendix at the end of the article.

• "Recommend a ryokan in Dotonbori, Osaka"
• "Where’s a good ryokan in Shinjuku, Tokyo?"
• "Where’s a good ryokan in Arashiyama, Kyoto?"
• "Recommend a hotel in Haeundae, Busan"

Data collection

1. Collected ChatGPT search queries. We recorded the web search queries ChatGPT ran for each prompt.
2. Analyzed Google’s top 10 results. We searched Google using the collected Korean queries and gathered the top 10 results.
3. Compared DRs. We collected the Ahrefs Domain Ratings of the sites in the results.

All figures reflect a single point in time in December 2025. The search queries and answers may vary each time the same question is run.

ChatGPT also searched in Korean

Language breakdown of search queries

We collected 346 search queries from 115 prompts.

Language	Number of queries	Share
English	225	65.0%
Korean	116	33.5%
Japanese	5	1.4%

Share of prompts that generated Korean search queries

At least one Korean search query appeared in 106 of the 110 prompts analyzed (96.4%). Although there were more English queries, Korean searches were also run for almost every prompt. This means Korean-language pages have a chance to be considered in search results for most questions.

Examples of actual search queries

Prompt	Query run
"Recommend a hotel in Matsuyama"	Matsuyama hotel recommendations
"Where’s a good hotel in Takamatsu?"	Takamatsu hotel recommendations
"Where’s a good hotel near Namba Station, Osaka?"	Hotel recommendations near Namba Station, Osaka

Korean OTAs can make Google’s top 10

DRs of major sites

Site	DR	Category
Booking.com	93	Global OTA
Naver Blog	93	Community
Agoda	89	Global OTA
Yanolja	71	Korean OTA
MyRealTrip	68	Korean OTA
Yeogi Eottae	55	Korean OTA

DRs are Ahrefs figures from December 2025.

DR breakdown of the top 10 results

We grouped the 1,060 Google top 10 results for 106 Korean search queries by DR range.

DR range	Number of results	Share	Competitiveness for Korean OTAs (DR 55–71)
Below 55	154	14.5%	High
55–68	190	17.9%	Possible to compete with content
69–80	254	24.0%	Possible with strong relevance to the question
Above 80	455	42.9%	Low

The table totals 1,053 results. The remaining 7 were not assigned to a DR range in the raw data.

The 344 results (32.5%) with a DR of 68 or lower are spots Korean OTAs can compete for. That is an average of 3.2 per query.

Yeogi Eottae example

At the time of the study, a Google search for "Osaka hotel recommendations" placed Yeogi Eottae at number 3, while Yanolja was not in the top 10. Yeogi Eottae’s DR (55) is lower than Yanolja’s (71). A city-specific recommendation page from Yeogi Eottae appeared in the results, while only accommodation search-results pages were found for Yanolja. This shows that even with a lower DR, a relevant content page can make the top 10.

A Google top-10 ranking does not guarantee a ChatGPT citation

Yeogi Eottae: ranked high, but rarely cited

Yeogi Eottae ranked number 3 on Google for "Osaka hotel recommendations," but it was rarely cited in the ChatGPT answers in this study.

The reason: a booking-interface structure

At the time of the study, Yeogi Eottae’s page had a typical booking-interface structure.

• The top of the page has date, price, and sorting filters.
• Hotel listings, star ratings, and long reviews follow.
• There are no sentences that lead with a conclusion, such as "Our recommended accommodations are A, B, and C."

To answer a recommendation question, a page needs sentences that give a conclusion and reasons. Filter screens and long lists of reviews offer few sentences that can be carried directly into an answer.

Room to compete with global OTAs

Agoda (DR 89) and Booking.com (DR 93) have DRs 18–38 points higher than the three Korean OTAs. However, the pages we examined on both sites also centered on booking interfaces.

Element	Agoda and Booking.com (at the time of the study)	Pages with recommendation content
Page format	Booking interface	Recommendation article
Sentences that can be carried into an answer	Few	Conclusive sentences such as "Our recommended accommodations are A, B, and C"

In this observation, whether a page had sentences that could be carried into an answer made a greater difference to citation than DR did.

Test case: TravelMorph

Test conditions

Item	Details
Domain	travelmorph.com (Anymorph test domain)
Domain Authority	0 (new domain)
Page creation	Anymorph pipeline
Test page	Couples’ Onsen Ryokan Recommendations in Arashiyama, Kyoto
Target question	"Ryokan with an onsen for couples in Arashiyama, Kyoto"

Results

Although the domain was new and had a Domain Authority of 0, the page was cited as a source in a ChatGPT answer 2 days after it went live.

Implications

This was a result from a single page and a single question, so it cannot be generalized. However, it shows that even a new domain with a low DR can be cited for long-tail questions when its page leads with a relevant conclusion. This test did not assess results for highly competitive queries such as "Osaka hotel recommendations."

Action items for Korean OTAs

Current state and goal

Item	Current	Goal
Page format	Booking interface	Recommendation content AI can read and reuse
Structure	Filters and listings	Summary with conclusions and descriptions for each accommodation
Content	User reviews listed	Editorial recommendations with stated criteria
Response to Korean search queries	Limited	Pages tailored to questions about specific regions and themes

Example recommendation page structure

Page element	Example
Title (H1)	5 Recommended Hotels in Osaka: Namba and Dotonbori
Summary (first paragraph)	Our recommended hotels in Osaka are A, B, C, D, and E. We selected them based on access to Dotonbori and price.
Subheading for each hotel (H2)	1. Hotel A, 2. Hotel B, and so on
Information for each hotel	Location: 3 minutes’ walk from Namba Station / Features: Dotonbori views, breakfast included / Price: from 150,000 won per night

Conclusion

1. ChatGPT also searched in Korean. Korean search queries appeared in 96.4% of the prompts analyzed.
2. There are top-10 spots Korean OTAs can compete for. Sites with a DR of 68 or lower made up 32.5% of the top results.
3. Content structure is needed to earn citations. A test page that led with its conclusion was cited 2 days after launch, even though it was on a new domain.

Even with lower DRs than global OTAs, Korean OTAs have opportunities to be cited in ChatGPT answers if they create recommendation pages for specific regions and themes. This study reflects a single point in time in December 2025, so after updating pages, the same questions should be measured repeatedly to assess the impact.

Appendix: Research data

Study overview

The study was conducted in December 2025 by Anymorph. It analyzed 115 prompts, 346 search queries, and 1,060 Google top 10 results.

115 test prompts

1. "Recommend a ryokan in Dotonbori, Osaka"
2. "Where’s a good hotel near Namba Station, Osaka?"
3. "Recommend accommodation in Shinsaibashi, Osaka"
4. "Recommend a hotel in Umeda, Osaka"
5. "Where’s a good place to stay in Tennoji, Osaka?"
6. "Where’s a good ryokan in Shinjuku, Tokyo?"
7. "Recommend an onsen hotel in Asakusa, Tokyo"
8. "Recommend a hotel in Shibuya, Tokyo"
9. "Where can I find good-value hotels in Ginza, Tokyo?"
10. "Recommend accommodation in Ikebukuro, Tokyo"
11. "Where’s a good hotel in Ueno, Tokyo?"
12. "Recommend a hotel in Roppongi, Tokyo"
13. "Recommend a hotel near Hakata Station, Fukuoka"
14. "Where’s a good ryokan in Tenjin, Fukuoka?"
15. "Recommend a hotel in Nakasu, Fukuoka"
16. "Recommend a ryokan in Gion, Kyoto"
17. "Where’s a good ryokan in Arashiyama, Kyoto?"
18. "Recommend a hotel near Kyoto Station"
19. "Recommend a hotel in Susukino, Sapporo"
20. "Where’s a good place to stay near Sapporo Station?"
21. "Recommend a ryokan in Osaka"
22. "Where can I find a good-value ryokan in Tokyo?"
23. "Recommend a ryokan in Kyoto"
24. "Where’s a good ryokan in Fukuoka?"
25. "Recommend an onsen hotel in Osaka"
26. "Where’s a good onsen hotel in Tokyo?"
27. "Are there any hotels in Osaka with tatami rooms?"
28. "Recommend a tatami hotel in Tokyo"
29. "Recommend onsen accommodation in Kyoto"
30. "Recommend a hotel in Nagoya"
31. "Where’s a good place to stay near Nagoya Station?"
32. "Recommend a resort in Okinawa"
33. "Where’s a good hotel in Naha, Okinawa?"
34. "Recommend a hotel in Hiroshima"
35. "Where’s a good ryokan in Kanazawa?"
36. "Recommend a ryokan in Hakone"
37. "Where’s a good ryokan with an open-air bath in Hakone?"
38. "Recommend an onsen ryokan in Beppu"
39. "Recommend a ryokan in Yufuin"
40. "Where’s a good ryokan at Arima Onsen?"
41. "Recommend a ryokan at Kusatsu Onsen"
42. "Where’s a good onsen hotel in Nikko?"
43. "Recommend a ryokan in Atami"
44. "Recommend an onsen ryokan in Noboribetsu"
45. "Where’s a good ryokan at Ginzan Onsen?"
46. "Recommend an onsen hotel in Shirahama"
47. "Recommend an onsen ryokan in Ibusuki"
48. "Where’s a good place to stay at Dogo Onsen?"
49. "Recommend a ryokan at Kinosaki Onsen"
50. "Recommend a ryokan in Takayama"
51. "Recommend a hotel on Khao San Road, Bangkok"
52. "Where’s a good hotel in Sukhumvit, Bangkok?"
53. "Recommend accommodation in Silom, Bangkok"
54. "Recommend a riverside hotel in Bangkok"
55. "Recommend a resort at My Khe Beach, Da Nang"
56. "Where’s a good place to stay in Hoi An, Da Nang?"
57. "Recommend a hotel near the Han River in Da Nang"
58. "Recommend a resort in Mactan, Cebu"
59. "Where’s a good hotel in Cebu City?"
60. "Recommend a villa in Ubud, Bali"
61. "Where’s a good hotel in Seminyak, Bali?"
62. "Recommend accommodation in Kuta, Bali"
63. "Recommend a resort in Nusa Dua, Bali"
64. "Recommend a hotel at Patong Beach, Phuket"
65. "Where’s a good resort at Karon Beach, Phuket?"
66. "Recommend a hotel in Marina Bay, Singapore"
67. "Where’s a good hotel in Orchard, Singapore?"
68. "Recommend a hotel in District 1, Ho Chi Minh City"
69. "Where’s a good place to stay in Hoan Kiem, Hanoi?"
70. "Recommend a hotel in Chiang Mai Old City"
71. "Recommend a resort in Jungmun, Jeju"
72. "Where’s a good pension in Aewol, Jeju?"
73. "Recommend a hotel in Seogwipo, Jeju"
74. "Where’s a good place to stay in Seongsan, Jeju?"
75. "Recommend a hotel in Haeundae, Busan"
76. "Where’s a good place to stay in Gwangalli, Busan?"
77. "Recommend a hotel in Seomyeon, Busan"
78. "Recommend a pension near Gyeongpo, Gangneung"
79. "Where’s a good place to stay in Jumunjin, Gangneung?"
80. "Where’s a good hanok stay in Gyeongju?"
81. "Recommend a hotel in Bomun Tourist Complex, Gyeongju"
82. "Recommend an ocean-view hotel in Yeosu"
83. "Where’s a good pension near Seoraksan, Sokcho?"
84. "Recommend accommodation in Jeonju Hanok Village"
85. "Where’s a good pension at Surfyy Beach, Yangyang?"
86. "Recommend a hotel near the Eiffel Tower in Paris"
87. "Where’s a good place to stay on the Champs-Élysées in Paris?"
88. "Recommend a hotel near Big Ben in London"
89. "Where’s a good place to stay near the Colosseum in Rome?"
90. "Recommend a hotel on La Rambla in Barcelona"
91. "Where’s a good place to stay in Prague Old Town?"
92. "Recommend a hotel near Dam Square in Amsterdam"
93. "Recommend a hotel in Times Square, New York"
94. "Where’s a good place to stay in Manhattan, New York?"
95. "Recommend a hotel in Santa Monica, LA"
96. "Where’s a good hotel on the Las Vegas Strip?"
97. "Recommend a hotel in Waikiki, Hawaii"
98. "Recommend a resort in Maui, Hawaii"
99. "Where’s a good hotel in Tumon, Guam?"
100. "Recommend a resort in Garapan, Saipan"
101. "Recommend a hotel near Universal Studios Japan in Osaka"
102. "Where’s a good place to stay near Tokyo Disneyland?"
103. "Recommend a hotel in Nara"
104. "Where’s a good hotel in Kobe?"
105. "Recommend a hotel in Yokohama"
106. "Where’s a good ryokan in Kamakura?"
107. "Recommend a hotel in Karuizawa"
108. "Where’s a good hotel in Hakodate?"
109. "Recommend a hotel in Otaru"
110. "Where’s a good hotel in Nagasaki?"
111. "Recommend a hotel in Kumamoto"
112. "Where’s a good hotel in Kagoshima?"
113. "Recommend a hotel in Matsuyama"
114. "Where’s a good hotel in Takamatsu?"
115. "Recommend a hotel in Sendai"

DRs of major domains

Domain	DR	Category
booking.com	93	Global OTA
tripadvisor.com	93	Reviews
expedia.com	91	Global OTA
agoda.com	89	Global OTA
japan.travel	81	Tourism board
japanican.com	74	Japanese OTA
matcha-jp.com	72	Travel guide
yanolja.com	71	Korean OTA
myrealtrip.com	68	Korean OTA
yeogi.com	55	Korean OTA