Hotel visibility in ChatGPT: 30 hotels in Nice measured
In short. On hotel visibility in ChatGPT and Google, our measurement of 30 independent hotels in Nice in November 2026 is clear. In 20 Google searches without a hotel name, 25 of the 30 websites appeared in the top results of none. In AI assistant answers that list their sources, 23 of the 30 websites were never cited. The causes are often simple: a home page titled “Home”, a site that opens in English, strengths nobody wrote down. The first fixes cost nothing.
Travellers rarely search for a hotel by name. They type “hotel Nice without a car” or ask an AI assistant which hotel is easy to reach by tram from the airport. For 30 independent hotels in Nice, we measured what Google and the assistants show for questions like these. This page presents the aggregated, anonymised results, the mistakes that keep coming back and the fixes a hotel can make itself. Each hotel in the measurement received its own details by email, and our page for hotels lets any of them receive those details free of charge.
How did we measure?
We studied 30 independent hotels in Nice in three groups of ten, on 4, 5 and 6 November 2026. For each hotel we ran 20 Google searches without a hotel name, on travellers’ themes: accessibility, breakfast, terraces, parking, travelling without a car. We also put travellers’ questions to several AI assistants with web search, including ChatGPT, Claude and Perplexity, with 9 answers per theme. Typical questions were which hotels are easy to reach by tram from the airport, which accept dogs, and which suit guests with reduced mobility. In each group, 120 answers listed their sources; those are the answers in which we count citations of hotel websites.
We count answers that name a hotel separately from answers that cite its website as a source. A hotel can be named thanks to a comparison site or a travel guide without its own site ever being read.
What do the Google searches show?
On the hotel’s own name, its website usually comes first. On the 20 searches without a name, the picture is almost uniform.
| Result across 20 searches without a name | Hotels | Share |
|---|---|---|
| Site in the top results of no search | 25 | 25 ÷ 30 = 83 % |
| Site in the top results of one search | 4 | 4 ÷ 30 = 13 % |
| Site in the top results of two searches | 1 | 1 ÷ 30 = 3 % |
Check the arithmetic: 25 + 4 + 1 is 30; 25 ÷ 30 is 0.83, 4 ÷ 30 is 0.13 and 1 ÷ 30 is 0.03. The few sites that did appear did so on one specific strength written down in plain words, such as a rooftop terrace or access without a car. On the hotel’s name, Google’s price comparison also showed, for several hotels, online travel agency offers before the hotel’s own site.
What do the AI assistants say?
Hotel websites are rarely read as sources. In the answers that list their sources, the assistants mostly cite comparison sites, online travel agencies and travel guides.
| Hotel website cited as a source | Hotels | Share |
|---|---|---|
| Never cited | 23 | 23 ÷ 30 = 77 % |
| Cited once | 5 | 5 ÷ 30 = 17 % |
| Cited three times | 1 | 1 ÷ 30 = 3 % |
| Cited five times | 1 | 1 ÷ 30 = 3 % |
Check the arithmetic: 23 + 5 + 1 + 1 is 30; 23 ÷ 30 is 0.77 and 5 ÷ 30 is 0.17. The site cited five times was also the most complete of the ten in its group. That is no accident: an assistant can only repeat what has been written.
The assistants are not only silent; they also get things wrong. Among the mistakes we recorded: a number of accessible rooms taken from a 2023 third-party listing and contradicted by the hotel’s own site; a 4-star hotel described as “mid-range”; a street name misspelt in several answers; a claim about accessible bathrooms that we could not find anywhere on the hotel’s pages. For more on this, read our guide on what to do when ChatGPT has wrong information about your business.
Why do real strengths stay invisible?
Because they are not written where assistants read. A hotel holding labels for dogs and for families was cited in none of the 9 answers about dogs and none of the 9 about families. By contrast, a hotel that clearly describes its rooftop terrace was cited in 7 of the 9 answers about terraces. The difference is not the quality of the hotel but the clarity of what its site and public sources say about it.
On the websites, the same weaknesses recur: a home page titled simply “Home”, with neither the hotel’s name nor Nice; a page with no description; a site that opens in English, with French kept separately. None of these is serious alone, but together they make the hotel hard to connect with the questions travellers ask.
Why is an independent hotel more exposed?
Because others speak about it in its place. Online travel agencies, comparison sites and guides publish thousands of pages about hotels in Nice, with descriptions, photos and reviews. When an assistant looks for an answer, it finds those pages first, and those are the pages it cites. The hotel is sometimes named correctly, but the traveller who clicks goes to the intermediary, not to the hotel’s site. A chain often has a rich central website that is updated regularly; an independent hotel usually has a shorter site, written once at opening. Advertising budget does not fix this. A site that itself answers, clearly and in the traveller’s language, the questions they ask does: access, accessibility, pets, families, parking, breakfast.
Which fixes cost nothing?
Four, in this order:
- Let the search crawler in. OpenAI’s documentation says that sites which opt out of OAI-SearchBot in robots.txt will not be shown in ChatGPT search answers, apart from plain navigational links. This setting is independent of GPTBot, which concerns training.
- Be indexed by Google. Google Search Central says there are no special requirements for AI Overviews: a page must be indexed and eligible to appear with a snippet. A page missing from the index will not enter those answers.
- Name the home page. A title with the hotel’s name and Nice, and a description that says in one sentence what sets it apart.
- Write down real strengths and correct old sources. Accessible rooms, pets, families, the terrace, tram access: a clear page for each, and updates to third-party listings that give old figures.
What next?
If you run a hotel in Nice and received our email, our page for hotels lets you receive the detailed findings free of charge: searches, answers and sources. If your hotel was not in the measurement, our guide to the AI visibility audit explains what an audit measures and what it costs, starting with a free express check. For the basics, see how to get recommended by ChatGPT.
What are the limits of this measurement?
It records what Google and the assistants showed on three dates, for given questions and conditions. Answers change with date, wording, language and location, so the same question asked next month may give a different list. Data we could not collect is not a zero, and a site that opens in English does not necessarily have a search problem on that basis alone. The measurement proves neither total invisibility nor lost bookings; it shows where to look first.
How do we work?
We agree the questions with the hotel, collect the answers with their dates, separate mentions from recommendations, trace the cited sources and write a report in plain language. The hotel sees the evidence behind every finding and decides what changes.
Send us the name of your hotel. We reply in writing within 48 hours.
Questions
How did you measure hotel visibility?
For each of the 30 hotels we ran 20 Google searches without a hotel name, the way travellers type them, and put travellers' questions to several AI assistants with web search, including ChatGPT, Claude and Perplexity. Each group of ten hotels produced 120 answers that list their sources.
Are the hotels named?
No. This page gives only aggregated figures and types of findings, with no hotel names. Each hotel received its own detailed findings by email.
What can a hotel do without paying anything?
Check that robots.txt does not block OAI-SearchBot, give the home page a title with the hotel's name and Nice, write its real strengths clearly on its own site, and correct old third-party listings that give wrong figures.
Does the measurement prove lost bookings?
No. It records what Google and the assistants showed on given dates for given questions. It proves neither total invisibility nor lost bookings, and answers change with date, wording and language.