What is in the dataset
The 576 answers cover Bali, Bangkok, Jakarta, London, New York, Paris, Singapore and Tokyo, with eight questions per city: best hotels, boutique, luxury, budget, family, business trip, couples, and one question in the local language. Together they name 5,156 hotel mentions across 1,471 different hotels and carry 3,945 source links from 595 websites.
| File | Rows | One row per |
|---|---|---|
| answers.csv | 576 | answer, with the assistant, city, question, run, hotels named and link counts |
| hotel_mentions.csv | 5,156 | hotel named in an answer, with its position, hotel ID and chain or independent label |
| hotels.csv | 1,471 | hotel per city, with mentions per assistant |
| sources.csv | 3,945 | link attached to an answer, with its domain, type and role |
| domains.csv | 595 | website, with its type |
A short script, metrics.py, recomputes every figure in our articles from these files.
Download all files: answers.csv, hotel_mentions.csv, hotels.csv, sources.csv, domains.csv, metrics.py and README.md.
How it was collected
ChatGPT and Gemini answers were collected from their web apps from the United States. DeepSeek supplied answers through its API with web search on. Each question ran three times on the same day, showing how much answers vary between runs alongside what they contain.
Hotel names and chain brands were extracted from each answer by a language model. Different spellings of the same hotel share one hotel ID, and each hotel carries the label it was given most often. Website types used the same method, with a few manual corrections. The labels give reliable totals, but check individual hotels before quoting them. The full answer text is not included.
What the data shows
The headline findings are written up in five articles: where AI assistants find hotels, why answers change between runs, independent hotels against chains, questions keyword tools miss and how answers differ by language. See also AI visibility for hotels.
Licence and credit
The dataset is published under CC BY 4.0. You can use, share and adapt it for any purpose, including commercial work, as long as you credit SearchSeal and link to this page. A suggested credit line is "Source: SearchSeal AI hotel answers dataset, October 2026".
FAQ
Can I use this data in a paper, article or product?
Yes, under CC BY 4.0 with credit and a link to this page.
Why is the answer text missing?
The dataset records the contents of answers (hotels, links and sources) rather than their full wording.
Will you update it?
Not on a fixed schedule. The date in the title is the collection date.