We asked ChatGPT 25 questions about hotels in Greece, in English, German and Greek. Not one of the 70 sources it cited was a hotel’s own website. Its recommendations come from magazine lists, travel blogs and OTAs, in other words from what other people write about your hotel.
Travellers increasingly use ChatGPT to build a shortlist before they open Booking.com or Google. If a hotel is not on that shortlist, it never gets compared. So we wanted to look at real data and answer two questions: which Greek hotels does ChatGPT recommend, and where does it get its information?
How we ran the study
On 30 October 2026 we sent ChatGPT 25 questions, phrased the way a traveller would type them. We used DataForSEO’s LLM Scraper, which records the answer as a user sees it on chatgpt.com, together with its sources, images and hotel cards.
| Market | Language | Questions | Examples |
|---|---|---|---|
| United Kingdom | English | 8 | Santorini, Mykonos, Crete for families, Rhodes all-inclusive, adults-only |
| United States | English | 9 | Paros, Milos, Halkidiki, Chania, budget Athens, family-run hotels |
| Germany | German | 4 | Crete, Rhodes, Corfu, Naxos |
| Greece | Greek | 4 | Halkidiki, Nafplio, Crete all-inclusive, Thessaloniki |
Each question ran once, with no user account and no history. ChatGPT does not always give the same answer, so individual hotel names can change from one run to the next. The patterns, however (where it gets its information and what kind of hotels it recommends), held steady across every market and language.
Which hotels appear
The 25 answers contained 202 hotel recommendations, covering 172 different hotels. The picture is highly fragmented: no hotel “dominates”, and the most frequent one, Domes Miramare in Corfu, appeared in 4 answers.
What does repeat is group names. Hotels that belong to groups and collections come up again and again, across destinations and languages:
| Group or collection (as it appears in the name) | Recommendations | Answers | Destinations |
|---|---|---|---|
| Domes | 8 | 7 | Corfu, Milos, Halkidiki, Zakynthos |
| Luxury Collection (Marriott) | 8 | 6 | Santorini, Corfu, Mykonos, Paros |
| Grecotel | 6 | 5 | Crete, Corfu |
| Electra | 4 | 3 | Athens, Kefalonia, Thessaloniki |
| Mitsis | 4 | 3 | Rhodes, Crete |
| Ikos, Atlantica | 3 each | 3 each | Halkidiki, Corfu, Rhodes, Zakynthos |
Two more observations:
- Language changes little. For family hotels in Crete, 5 of the 7 hotels were the same in the German and the Greek answer (Creta Maris, Stella Palace, Lyttos Beach, Nana Golden Beach, Fodele Beach).
- Small hotels do get in, but on narrower questions. General questions (“best hotels in Santorini”) returned only five-star hotels in Oia and Imerovigli. More specific ones (“family-run hotels”, “a weekend in Nafplio”, “Chania old town”) brought up small boutique and family-run hotels, such as Anemomilos on Folegandros, Amalthia in Pelion and Carpe Diem in Nafplio.
Where ChatGPT gets its information
In 18 of the 25 answers ChatGPT showed sources, 70 citations in total. The most frequent were Tripadvisor (in 6 answers), Condé Nast Traveller (in 5), santorinidave.com (in 4), The Hotel Guru and The Luxury Editor (in 3 each).
Three things stand out:
- One blog covers all of Greece. santorinidave.com, an English-language travel site, was cited for Santorini, Paros, Chania and Halkidiki.
- OTAs matter, but they do not dominate. Booking.com, Tripadvisor, Expedia, Hotels.com and HolidayCheck were cited as a source in 11 of the 25 answers. In the images, though, they are everywhere: in 21 of the 25 answers at least one hotel photo came from an OTA.
- In Greek, the sources were English. In the 4 questions we asked in Greek, not one citation was a Greek-language site. Two of them had no sources at all.
In 7 of the 25 answers there were no sources at all. There, the traveller cannot see where the recommendations come from, and the reputation a hotel has built over previous years counts even more.
Why these hotels
The hotels ChatGPT recommends share five traits. None of them is a technical trick. All of them come down to how clearly and how often other people write about the hotel.
- They are already on “best hotels” lists. To a large extent ChatGPT summarises the lists it reads. For Santorini it said so explicitly: its picks were based on Condé Nast Traveller, Time Out and specialist guides. 42 of the 70 sources had “2026” or “2027” in their title, so recent lists are what count most.
- They have a clear profile. Answers are organised by type of traveller and by area, with labels such as “best for couples”, “adults-only”, “all-inclusive with an aqua park” or “Oia, five-star”. A hotel that is described the same way everywhere slots easily into a category.
- They have many good reviews. In Mykonos every recommendation came with a Tripadvisor rating. In Chania, Booking.com scores were quoted (e.g. Hanim 9.8, La Maison Ottomane 9.7). In Paros, Saint Andrea appeared with 4.9 out of 5 from 2,293 reviews.
- They belong to groups or collections. A group with many hotels appears on more lists and in more destinations, so names like Domes, Grecotel, Mitsis and Luxury Collection keep coming back.
- They appear in authoritative directories. Michelin Guide pages were used for the photos in 6 of the 25 answers, and a 500 Best Hotels Greece award was cited as a source.
The official website matters, just somewhere else. In 15 of the 25 answers ChatGPT showed a map card with the recommended hotels. Of the 106 entries in those cards, 105 linked to the website of the hotel or its group, none to an OTA, and one had no website. That is where a direct booking can happen, as long as the website is correctly linked to the hotel’s profile.
At the end of every answer ChatGPT asked the user about budget, who is travelling and when. Its next recommendation therefore narrows down to hotels that fit a specific profile. The more clearly yours is described, the more likely you are to be in it.
What a Greek hotel can do
The priority is to get into the sources ChatGPT reads, not to “optimise” your website for AI. In this order:
- Find out which sources matter for your destination. Ask ChatGPT 10–15 questions the way your guests would, in the languages of your markets (e.g. “best adults-only hotels in Naxos”, “Familienhotel Kreta mit Aquapark”). Note which hotels appear and which sites are cited. That is your target list.
- Get onto those lists. For most destinations the sources are few: one or two local blogs (such as naxosinsider.com or guidetomilos.com), a hotel curator (The Hotel Guru, The Luxury Editor, itsallinclusive.com) and a travel magazine. Offer them a stay, photos and whatever they need to write about you. English and German media pay off more than Greek ones, even for questions asked in Greek.
- Choose one clear profile and use it everywhere. If you are an “adults-only boutique hotel in Agia Anna” or a “family all-inclusive with an aqua park in Hersonissos”, your website, Google Business Profile, Booking.com page and Tripadvisor listing should all say so in the same words. ChatGPT groups hotels by exactly these labels.
- Work on your reviews on Google, Tripadvisor and Booking.com. Actively ask every guest for a review and reply to all of them. Ratings show up inside the answers themselves.
- Check your Google Business Profile and the link to your website. Map cards point to the website you list there. If the link is wrong or missing, you lose the most direct route to a direct booking.
- Make sure your website can be read. Check that robots.txt or your firewall (e.g. Cloudflare) does not block ChatGPT’s crawler (
OAI-SearchBot). You need pages in English and German that describe your type of hotel, your area and your facilities in plain, specific language. - Measure every month. Run the same questions every month and record whether you appear, in what position and through which source. Lists are refreshed every year, and pre-season (November to March) is the right time to get into the new ones.
Frequently asked questions
Can I pay to have ChatGPT recommend my hotel?
None of the 25 answers we recorded contained an ad. A place in these answers is earned through what others write about your hotel.
Is this worth it for a small hotel or a villa?
Yes, but on more specific questions. General questions are won by five-star hotels and groups. Questions about family-run hotels, a weekend in Nafplio or Chania’s old town brought up hotels with only a few rooms.
Does my website’s SEO help?
Indirectly. Your website was not cited as a source in any answer, but it is where the map cards lead, and the basis for whatever bloggers and journalists will write about you.
How quickly do results change?
When ChatGPT searches the web, a new list or a change in your reviews can show up quickly, because it reads pages at the moment of the question. In 7 of the 25 answers, however, no source appeared at all. There, change is slower, because the reputation built over previous years carries more weight.
Does the same apply to Gemini and Google AI Overviews?
This study covered ChatGPT only. The principles are similar, but the sources may differ, so it is worth checking them separately.
Want to see where your own hotel stands?
At Quadrum Digital we run the same analysis for your hotel. We put your markets’ questions to ChatGPT, find which competitors appear and through which sources, and give you a list of actions to get into those sources before the season.
Book an AI visibility audit · See our AI SEO service
Data source: 25 questions to ChatGPT via DataForSEO LLM Scraper, 30 October 2026. Analysis: Quadrum Digital.