How AI chooses hotels,
one city at a time
Each Stiplo study takes every registered hotel in one city, asks the AI assistants the questions travellers really ask, and checks every answer against every property. The method stays the same. The question changes with each city.
Study 04 · Next city · To be announced
Be first to read the next city
One email when a study is published, with the headline findings and the full report. No weekly letter, nothing else in between.
Three cities, three questions
Each study picks up where the last one stopped
03
Bangkok · November 2026
It took the method outside Europe
The third study asked as travellers from four markets, the UK, the US, India and Japan (in Japanese), and tested whether AI's shortlist for a city holds up once the travellers come from different parts of the world.
Read the Bangkok study- 43%of the 1,260 hotels measured were never named in 24,480 answers, by any assistant, for any market
- 94%of all mentions go to the top fifth of hotels (252 hotels)
- 0 / 7times an AI agent asked to book a hotel went to a booking site: 12 hotels AI names often, against 12 it never names. A separate, small Stiplo test.
How the studies differ
| Published | Properties measured | AI answers | Asked | The question | Never named by AI | Independents and chains | Where the link goes | |
|---|---|---|---|---|---|---|---|---|
| Bangkok Study 03 | November 2026, with NoVacancy Asia and GAIN | 1,260 hotels from booking listings, 3 to 5 star or unrated | 25,600 | For 4 markets: the UK, the US and India in English, Japan in Japanese | Who does AI recommend in Bangkok, and who does it skip | 43%, more than 2 in 5 | 47% of independents named at least once, against 85% of branded hotels | ChatGPT's hotel cards go to the hotel's own site; Perplexity cites a booking site in at least 4 of 5 answers |
| London Study 02 | November 2026, with WTM London | 1,485 registered hotel-class properties | 39,840 | From 3 countries: the UK, the US, Australia | Once AI names a hotel, who gets the click | 1 in 7 | 3 times as likely, and no gap like for like | ChatGPT to the hotel, Gemini and AI Overviews to Google first |
| Barcelona Study 01 | 21 July 2026 | 659 registered hotels, hostals and guesthouses | 27,360 | In 3 languages: English, Spanish, Catalan | Who does AI name, and who does it leave out | 22%, about 1 in 5 | Independents 8 times as likely to be invisible | The hotel's own site is cited 7.3% of the time |
Scroll sideways for every column
What stays the same in every city
The whole register
Every registered property in the city is checked. Not a sample.
Real traveller questions
The questions people really ask an assistant before they book.
Asked 20 times each
AI answers change from one run to the next, so every question is repeated.
Four assistants
ChatGPT, Gemini, Google AI Overviews and Perplexity.
The studies measure whole cities, a Snapshot looks at yours
We need nothing but your web address. Running a group? We scan the whole portfolio.
In your Snapshot
- 01How often AI recommends you
- 02Visibility by assistant and topic
- 03Your top Website Health findings
- 04From insight to action


