Same City, Different Language, Different Hotels
We asked AI assistants for hotels in Paris and Tokyo in French, Japanese and English. Only about 1 in 5 hotels matched across languages. Here is why it happens.
Does the question's language change which hotels AI picks?
Most hotels assess AI visibility in English, although many guests ask in another language. In our study of 576 AI hotel answers, five cities had one question in the local language alongside the English "best hotels in [city]". To separate language effects from normal run-to-run change, we compared local-language answers with English answers, then compared English answers with repeat English answers for overlap.
Most hotels assess AI visibility in English, although many guests ask in another language. In our study of 576 AI hotel answers, five cities had one question in the local language alongside the English "best hotels in [city]". To separate language effects from normal run-to-run change, we compared local-language answers with English answers, then compared English answers with repeat English answers for overlap.
| City | Local-language question | Overlap with English answer | English asked again |
|---|---|---|---|
| Paris | meilleur hôtel à Paris | 19% | 61% |
| Tokyo | 東京 おすすめ ホテル | 19% | 58% |
| Bangkok | โรงแรมที่ดีที่สุดในกรุงเทพ | 34% | 58% |
| Bali | rekomendasi hotel terbaik di Bali | 48% | 68% |
| Jakarta | rekomendasi hotel terbaik di Jakarta | 46% | 48% |
In Paris and Tokyo, the local-language answers were much further from the English ones than the English answers were from each other. Bangkok had a smaller gap. In Jakarta and Bali the gap fell within normal run-to-run change, so we can't attribute it to language.
What changed in Tokyo
In English, ChatGPT's "best hotels in Tokyo" was a luxury list of Aman Tokyo, Palace Hotel, Park Hyatt, Hoshinoya and Mandarin Oriental. Asked in Japanese, it also recommended practical mid-range hotels, including Hotel Sunroute Plaza Shinjuku, Hotel Ryumeikan, MIMARU and APA Hotel Asakusa, with The Tokyo Station Hotel alongside them.
Some of this comes from the wording, since the Japanese question asks for "recommended" hotels rather than the "best", which is the usual way to ask in Japanese. Whatever caused it, guests asking in Japanese got a different kind of shortlist.
Why this happens
The likeliest reason is that assistants search in the language of the question, so a Japanese question can pull Japanese guides and booking pages and a French question French ones. Different sources produce a different shortlist.
What this means for hotels
- Check the languages your guests use. If many guests come from Japan, Korea or Indonesia, check AI answers in those languages alongside English.
- Get listed in local-language guides and on local booking sites. A hotel found only in English sources will likely appear mostly in English answers.
- Translate the facts along with the marketing. Room types, location and who the hotel suits matter in each language.
How SearchSeal helps
SearchSeal lets you track questions in the languages your guests use, so a hotel can follow its English, Japanese and Indonesian questions side by side. See AI visibility for hotels.
Method and limits
- One local-language question per city, three runs each on ChatGPT, Gemini and DeepSeek, compared with the same assistant's English "best hotels in [city]" answers. Overlap is the hotels named in both answers divided by the hotels named in either. 21 October 2026.
- One question per city is a small sample, so treat this as a reason to check rather than a rule.
- The local questions use natural phrasing in each language rather than word-for-word translations, so some of the difference comes from wording.
- ChatGPT and Gemini answers were collected from the United States. Guests asking from Tokyo or Jakarta may get different results.
Frequently asked questions
Does the language of the question change which hotels AI recommends?
In Paris and Tokyo, the local-language answers were much further from the English ones than the English answers were from each other. Bangkok had a smaller gap. In Jakarta and Bali the gap fell within normal run-to-run change, so we can't attribute it to language.
Why do local-language answers differ?
The likeliest reason is that assistants search in the language of the question, so a Japanese question can pull Japanese guides and booking pages and a French question French ones. Different sources produce a different shortlist.