How Travelers Actually Use ChatGPT, Gemini & Perplexity to Pick a Hotel

Here’s a habit that snuck up on the whole travel industry while everyone was busy arguing about Google rankings: people stopped starting their hotel search on Google.

Or rather, a growing slice of them did. Instead of typing “boutique hotels Asheville” into a search box and wading through ten blue links and a wall of OTA ads, they open ChatGPT and type something like, “I’m going to Asheville for a long weekend with my partner, we want somewhere walkable, romantic, and not a chain, under $300 a night. What do you suggest?”

And the assistant just… answers. With names. A little blurb on each. A reason. Sometimes a comparison table. It reads like advice from a friend who happens to have read the entire internet.

If you run an independent or boutique hotel, that shift should get your attention, because the moment a traveler forms their shortlist is the moment the booking decision really begins. And increasingly, that moment is happening inside an AI conversation you’ve never seen. Let’s walk through exactly how travelers are using these tools, with real prompts and the kind of answers they produce, so you can see precisely where you need to show up.

First, the lay of the land

There are three assistants doing most of the heavy lifting in trip planning right now, and they each behave a little differently.

• ChatGPT is the big one by raw reach. It’s conversational, it remembers context across a long planning session, and travelers treat it like a knowledgeable concierge they can keep pestering.
• Perplexity is the researcher’s tool. It leans hard into citations and links, so it attracts the kind of traveler who wants to verify things and click through to sources. Great news for hotels, because Perplexity answers come with footnotes.
• Gemini lives inside Google’s gravity well. It powers a lot of what shows up in AI Overviews at the top of search results, so being visible there overlaps with being visible in classic Google.

Why does the difference matter to you? Because a traveler rarely uses just one. They’ll brainstorm in ChatGPT, fact-check a claim in Perplexity, then Google a specific hotel name and land on a Gemini-flavored AI Overview. Your job isn’t to win one box. It’s to be a consistent, accurate, quotable presence across the whole messy journey. We dug into the mechanics of that in AEO vs GEO vs SEO for hotels if you want the taxonomy.

The five things travelers actually do with AI

When you watch how trip planning unfolds, AI assistants get used at five distinct stages. Each one is a chance to be present, and each one rewards different things.

1. Discovery: “Where should I even go?”

This is the top of the funnel, vague and dreamy. The traveler doesn’t have a hotel in mind yet. They might not have a city in mind.

“We want a warm-weather getaway in March, somewhere walkable with good food, not too touristy, flying out of Atlanta. Where should we go and where should we stay?”

The assistant responds by suggesting a few destinations and, crucially, it often name-drops a couple of standout places to stay in each. If your property is the kind of place a model associates with “characterful, walkable, food-forward town,” you can surface here before the traveler has even committed to a location. That’s about as early as the booking decision gets.

What gets you mentioned at this stage isn’t a discount code. It’s being clearly, unmistakably about something. The 1880s mill conversion. The 14-room inn with the chef who forages. The hotel with the rooftop that locals actually go to. Models reward distinctiveness because distinctiveness is what makes a recommendation useful.

2. Shortlisting: “Give me five options”

Now the traveler has a destination and wants candidates. This is the prompt that most directly decides whether you’re in the running.

“Best independent boutique hotels in Savannah’s historic district for a couples trip, walking distance to restaurants, with character. Give me 5 with a quick note on each.”

The answer is a tidy list of five names, each with a one-line pitch. If you’re on it, congratulations, you made the shortlist a human will actually consider. If you’re not, you’re invisible at the exact moment the field narrows from “everything” to “five.” There is no page two in a chatbot. There’s no scrolling to find you. You’re either named or you don’t exist.

This is why we keep telling hoteliers that AI invisibility is more brutal than a bad Google ranking. A mediocre Google position still gets seen. Being absent from a five-item AI list is a clean omission. We wrote a whole piece on this exact fear: is your hotel invisible to ChatGPT?

3. Comparing: “Which of these is better for me?”

Once a shortlist exists, travelers pit options against each other. This is where structured, factual information about your property earns its keep.

“Compare The Foundry Hotel and The Restoration for a foodie couple, focus on location, vibe, breakfast, and price.”

Assistants love to answer this with a comparison table. Here’s the kind of thing they generate, and notice how much of it is just facts about the property that either exist clearly online or don’t:

That table is illustrative, not real data, so don’t quote it. But the lesson is real: the assistant filled those cells from information it could find and trust. If your breakfast policy, walk times, and vibe aren’t stated plainly somewhere a model can read them, you show up as a row full of “unclear” or, worse, you get facts wrong in a way that loses the booking. Making your details machine-readable is exactly what structured data to make your hotel quotable to AI is about.

4. Itinerary-building: “Plan my whole trip”

This is the sleeper stage everyone underrates. Travelers increasingly ask AI to build the entire trip, and the hotel gets woven in as the home base.

“Plan a 3-day Charleston itinerary for two, staying somewhere walkable downtown. Include where to stay, eat, and what to do each day.”

When the assistant builds this, it picks a hotel and then anchors the whole plan around it: breakfast here, walk ten minutes to this gallery, dinner there. If you’re the chosen base, you’re not just a booking, you’re the center of gravity for the entire trip. That’s a powerful position, and it goes disproportionately to hotels with clear location signals and a strong sense of what’s nearby. Helping the model understand your neighborhood matters as much as describing your rooms.

5. Interrogating: “Tell me about the actual details”

Right before booking, travelers get specific and a little anxious. They ask the questions that close or kill a sale.

“Does The Foundry Hotel have free parking, a pool, and can I check in early? Is it walkable to the City Market?”

These are exactly the questions your front desk fields by phone all day. When an assistant answers them confidently and correctly, it removes friction at the worst possible moment to have friction. When it answers them wrong, or shrugs, the traveler bounces to a property whose details are crystal clear. The difference is almost never the amenity itself. It’s whether the fact is stated somewhere a model can find and trust it. This is the whole game behind getting cited in Google AI Overviews, and it’s why a structured llms.txt for hotels is becoming a sensible move.

So where does the booking decision actually start now?

Here’s the uncomfortable reframe. For a long time, the mental model was: traveler searches, sees options, clicks, books. The booking “started” on a search results page or an OTA listing.

That starting line has moved earlier, and somewhere you can’t see. By the time many travelers reach an OTA or your booking engine today, they’ve already had a conversation with an AI that shaped which two or three names they’re even considering. The OTA isn’t where they decided anymore. It’s increasingly just where they transacted, after the deciding happened in a chat window.

This matters enormously for the OTA question, so let’s be precise about it.

The honest OTA angle

We are not going to tell you AI search lets you fire the OTAs. It doesn’t, and anyone promising that is selling you something. The OTAs have enormous distribution, billions in marketing spend, and a place in traveler habits that a single property cannot replace. Booking.com and Expedia will keep sending you guests, and you’ll keep paying roughly 15-25% in commissions for that reach. That’s the deal, and pretending otherwise is how hoteliers get burned.

What AI visibility can do is more modest and more honest: it lets you show up earlier in that planning conversation, on your own terms, with your own name and your own facts, before the OTA listing is the only thing the traveler sees. When you’re present at the discovery and shortlisting stages, a meaningful number of travelers will arrive at your direct site already convinced, rather than meeting you for the first time inside a Booking.com search where you’re one tile among forty.

That’s the lever: reduce your OTA dependence, claw back some margin, and win a healthier booking mix by being part of the decision before the decision reaches the OTA. Not escape. Not “beat the OTAs.” Just a few more direct bookings a week and a little less of every reservation going to commission.

What this means for you, practically

You don’t need to become an AI engineer. You need to make sure that when travelers run these five kinds of prompts about your market, your hotel shows up accurately and with a clear path to your own site. The work breaks down into a few honest moves:

1. Audit what the assistants currently say about you. Open ChatGPT, Gemini, and Perplexity and run the exact prompts above with your town and property names. Note where you’re absent, where you’re mentioned, and where you’re flat wrong. We turned this into a repeatable process in audit what ChatGPT says about your hotel.
2. Make your core facts unmistakable. Parking, breakfast, pet policy, walk times, vibe, what’s nearby. Stated plainly, in your own words, on your own site. Models can’t quote what you never wrote down.
3. Be distinctive on purpose. “Nice hotel in town” gets you nowhere. “The restored 1890s firehouse with the courtyard bar” gets you named.
4. Earn third-party mentions. Assistants trust corroboration. Getting your name into local guides, press, and credible roundups feeds the citations these models lean on. That’s the logic behind the GEO trifecta of schema, entity, and citations.
5. Understand the full traveler journey. The deeper version of this whole topic lives in how travelers use AI to pick hotels… which is this post, so instead go read about structured data and our AI visibility service if you’d rather we handle it.

The big shift is simple to say and hard to ignore: travelers now form opinions about your hotel in conversations you don’t control and can’t see. The hotels that win the next few years aren’t the ones with the cleverest ad spend. They’re the ones a model can describe accurately, distinctively, and with a clear nudge toward booking direct.

Ready to find out what the machines say about you?

If you have ninety seconds, go ask ChatGPT for the best boutique hotels in your town right now and see whether you make the list. If you don’t love the answer, that’s exactly the problem we fix. Book a free intro call and we’ll run a live audit of how you show up across ChatGPT, Gemini, and Perplexity, then map out how to get into more of those answers and claw back more direct bookings. Or read more about our AI visibility work first. Either way, the conversation about your hotel is already happening. Let’s make sure you’re in it.