What Does ChatGPT Say About Your Hotel or Destination? How to Check AI Visibility
A method any destination, hotel or attraction can follow: which questions to ask ChatGPT, Gemini, Claude and Perplexity, in which languages, and what to record.
To find out what ChatGPT says about your hotel, destination or attraction, ask it the questions your travelers ask, and write down what comes back. Use four assistants (ChatGPT, Gemini, Claude, Perplexity), in the languages your visitors speak, several times each. For every answer, record whether you are mentioned, whether the facts match your official ones, which sources are cited, and who is recommended instead.
That is the whole method. The rest of this guide is how to do it so the result means something, and what to look for in the answers.
How do I check what ChatGPT says about my hotel or destination?
Set aside two hours and follow five steps.
- Write down your official facts first. Ten lines are enough: what you are, where exactly, what you offer, what makes you different, opening dates, the practical details travelers ask about (parking, pets, accessibility, how to get there). This is the reference you will grade every answer against. Without it, you are judging answers by feel.
- Write ten to fifteen questions in three groups (see the next section): questions that name you, questions about your market that don’t, and questions about what travelers think of you.
- Ask each question in a fresh conversation. Log out, or use a temporary chat, so your own history doesn’t shape the answer. Leave web search on: that is how most travelers use these assistants today, and it is the only way to see which sources they cite.
- Ask each question three times. Answers vary from one run to the next (more on that below). Three runs per question per assistant is the minimum to tell a pattern from chance.
- Record every answer in one spreadsheet, one row per answer, with the columns in the table further down.
Which questions should I ask?
Ask the questions travelers ask, not the ones you would like to be asked. They fall into three groups, and each one tells you something different.
Questions that name you. “Tell me about Hôtel du Lac in Annecy.” “Is Camping des Pins good for families?” “What is there to do in the Vaucluse in spring?” These show what AI knows about you: whether it describes the right place, with the right facts.
Questions about your market that don’t name you. “Where should we stay on Lake Annecy with a dog?” “Best family campsites in the south of the Ardèche.” “Where to go in Provence for a long weekend?” This is how most trips start: travelers rarely know your name yet. These questions show whether AI picks you, and who it picks instead. For a destination, the market is the travel area one level up: an audit of the Vaucluse asks about Provence.
Questions about how travelers see you. “What do guests complain about at Hôtel du Lac?” “Is Annecy worth it in August or is it too crowded?” These show what AI repeats from reviews and forums: the strengths it credits you with, and the reservations it passes on.
Six brand questions, six market questions and three perception questions make a good first set.
Which AI assistants and languages should I check?
Check the four assistants travelers use for trip planning: ChatGPT, Gemini, Claude and Perplexity. They don’t read the same sources and they don’t agree. A destination can be well described by one and confused with a namesake by another.
Ask in every language your visitors speak, not only yours. The answer about a French campsite in English is often built from different sources than the answer in French: an English-language travel guide or an OTA listing, where the French answer quotes the tourist office. If a third of your guests are German or Dutch, the German and Dutch answers are a third of your AI Visibility.
What should I record for each answer?
| What to record | What to look for | What it tells you |
|---|---|---|
| Mentioned or not | Are you named at all? On market questions, in what position and how often across runs? | Whether AI considers you an option. Count the share of answers that name you, not your rank in one answer. |
| Accuracy against your official facts | Wrong location, old prices or dates, amenities you don’t have, claims you never made, a namesake’s details | Whether AI describes you or a guess of you |
| Sources cited | Your own website, an OTA, Tripadvisor, Wikipedia, a travel guide, a forum | Who speaks for you. A correct answer that cites only third parties still sends the traveler elsewhere. |
| Rivals named | Which hotels, sites or destinations AI recommends instead of you, or next to you | Your real competitive set in AI answers, which is often not the one you benchmark |
| Tone and reservations | “Crowded in summer”, “dated rooms”, “hard to reach without a car” | What AI holds against you, and whether it is still true |
Two numbers come out of this sheet. Your mention rate is the share of market answers that name you. Your citation rate is the share of answers that cite your own website. Accuracy is a third, but you will read it line by line rather than count it.
What does a wrong AI answer usually look like?
Rarely a dramatic invention. Most errors we see are ordinary facts, slightly wrong, repeated with confidence. In the pilot audits we ran for destinations, hotels and campsites in September and October 2026, the errors fell into a few recurring patterns:
- A namesake. Two hotels share a name, and the answer mixes them: the other one’s amenities, the other one’s village.
- A stale number. A campsite’s pitch count or a hotel’s room count from an old page or an old listing, still quoted after the change.
- A wrong practical fact. A train time to the capital twice as long as the real one, given as fact.
- The wrong scope. A question about one destination answered for the whole region around it.
- A flattering claim that isn’t true. A superlative (“the largest in Europe”) the destination never made and cannot stand behind.
Each one has a cause on the web, usually on your own site: a fact shown only in an image, a page nobody updated, no exact coordinates, no link to your official profiles. Our guide to what hotels should fix on their website goes through those causes one by one.
Why does the answer change every time I ask?
Because the assistants generate each answer afresh, and that is why you ask three times. In January 2026, SparkToro had 600 volunteers run the same brand-recommendation prompts 2,961 times through ChatGPT, Claude and Google’s AI. They found less than a 1 in 100 chance that ChatGPT or Google’s AI returns the same list of brands twice, and roughly 1 in 1,000 that it returns it in the same order (SparkToro, 27 January 2026).
What does hold steady is how often a brand appears across many runs. So don’t read a single answer as your position. Read the share of answers that mention you, and watch it move month after month.
Is AI Visibility the same as AI Readiness?
No, and keeping them apart saves you from fixing the wrong thing. AI Visibility is the outcome: what the assistants say, whether it is accurate, whether they cite you. AI Readiness is one cause: whether your website can be read and understood by AI in the first place. Visibility varies by assistant, language and run. Readiness is deterministic: the same site gets the same checks.
Most travel websites are not easy for AI to read yet. In the October 2026 edition of our AI Readiness benchmark, official destination websites average 30.0 out of 100 (906 sites) and hotel websites 27.6 (2,095 sites). When your website is hard to read, AI takes your facts from the sources that are easier to read, and the answer becomes theirs. You can check your website’s AI Readiness for free, without signing up, and read how the score works.
Together, the two audits are what we call AI Presence: how AI understands, represents and interacts with your brand. SEO got you into the list; AI Presence gets you into the answer.
How does the Globetrotters AI Visibility audit work?
It runs the method above for you, with travel questions written for your type of place, and reads every answer in three lenses:
- Brand: how AI describes you when travelers name you, graded against your official narrative with Narrative Alignment (A to F), and which sources the answers cite, starting with your own website.
- Market: the questions travelers ask about your market without naming you, and the destinations or places AI recommends instead, sorted between your own territory and your competitors.
- Perception: what AI says travelers like, dislike and hold against you.
| Audit (free) | Monitor (€29/month) | |
|---|---|---|
| Lenses | Brand, a snapshot at signup | Brand, Market and Perception |
| Assistants | ChatGPT and Gemini | ChatGPT, Gemini, Claude and Perplexity |
| How often | Once | A new report every month, with history |
| Languages | One | One, more as an option |
| Install on your website | Nothing | Nothing |
Prices excluding VAT, as listed on our pricing page on 2 November 2026. Monitor comes with a 14-day free trial, no card required.
Monitor reads the assistants from the outside: no plugin, no code, no change to your site. When you are ready to change what the answers say, Publish puts your official facts on your own domain in the formats AI agents read. If you are comparing tools, our guide to AI visibility tools for travel brands sets the main options side by side.
Start with the free snapshot. Create your free Audit account and see what ChatGPT and Gemini say about your destination, hotel or attraction today, and which sources they cite.
Sources
- SparkToro, AIs are highly inconsistent when recommending brands or products, Rand Fishkin with Patrick O’Donnell (Gumshoe.ai, which sells AI tracking tools), 27 January 2026.
- Globetrotters.ai, AI Readiness statistics for travel brands, October 2026 edition: 906 official destination websites and 2,095 hotel websites in 21 countries.
- Globetrotters.ai, AI Visibility pilot audits of destinations, hotels and campsites, September and October 2026: the error patterns above, aggregated, with no organization named.
