Travelers have started asking AI assistants for hotel recommendations. A plain-language look at how answer engines pick hotels, why most independents are invisible to them, and what to do about it now.
A note for hotel marketers watching a new channel form — and wondering whether it’s real.
A traveler types into an AI assistant: “Quiet 4-star near the old town in Vienna, good breakfast, under €180 — what would you pick?” The assistant doesn’t show ads. It reads the web, picks a few sources it trusts, and answers — two or three hotels, a sentence of reasoning each, and links.
If your hotel is in that answer, you just received the warmest kind of referral: a recommendation, with your direct link, in a conversation with zero competing banners. If it isn’t — and for most independent hotels, it isn’t — you didn’t lose the booking. You were never considered.
This is a new distribution channel forming in plain sight. It’s early, it’s small, and it’s worth understanding now, while the cost of showing up is low.
How an AI assistant “sees” a hotel
An answer engine doesn’t browse the way a guest does. It reads pages as text and structure, and it builds answers from facts it can verify and quote. That has three practical consequences.
It can’t operate your booking widget. A hotel website where the price appears only after selecting dates inside an interactive booking flow is, to a machine reader, a page with no price on it. The photos may be beautiful; the facts are invisible.
It prefers sources it can trust. Consistent facts across the web, a named source for the guest score, structured data that says this is a hotel, this is its address, this is its rating — these are what let an assistant include you in an answer without risking being wrong. Assistants are cautious by design; unverifiable claims get skipped, not repeated.
It follows links it’s given. When the answer includes a booking link, the assistant uses the one it found. If the only bookable link it can find for your property is an OTA page, the recommendation — even when you win it — carries a commission.
The asymmetry that makes this worth doing now
Let’s be honest about the size: AI-assistant referrals are a small share of hotel bookings today, and anyone quoting precise percentages for 2028 is guessing. We won’t.
What can be said with confidence is that the cost structure is asymmetric. Making your hotel readable to answer engines is cheap, one-time-ish work that also improves ordinary search. Being unreadable costs nothing today — and an unknown, growing amount every quarter after. And early sources compound: assistants keep drawing on sources that answered well before. The hotels that are citable early become the default citations later.
You don’t have to believe this channel becomes huge. You only have to notice that the bet is priced like insurance, not like a gamble.
What a hotel can do now
Vendor-neutral, in rough order of value:
1. Have one page where the facts live in plain, readable form. Name, address, amenities, a real description, the guest score with its source named — as text a machine can read, not only as images or a PDF fact sheet. Server-rendered, no interaction required.
2. Add structured data. Schema.org hotel markup (the machine-readable label that says “hotel, address, rating, price”) is the difference between an assistant guessing what your page is and knowing. Your web agency can add it; it’s standard work.
3. Make the direct price visible outside the booking engine. A live “from” rate on a readable page gives assistants a fact to quote — and a reason to link you rather than the OTA that does publish a price.
4. Check the signposts. Sitemap current, robots.txt not blocking AI crawlers (several hotels block them by accident, inheriting a default), and an llms.txt if you want to be explicit about what’s where.
5. Measure it. Two habits: check your server logs (or a crawler log, if your tooling has one) for visits from AI and search bots — and once a month, ask the major assistants about hotels in your city and see whether you appear, and which source they cite when you do.
What we do here — disclosure
We practice this on our own site — it’s why this blog has an llms.txt and explicitly welcomes AI crawlers. And we build it for hotels: the Discovery module publishes exactly the kind of page described above for each hotel — live direct rate from your own system, guest score with the source named, the full entity, and a commission-free direct-booking link — plus a log of every visit from known AI and search crawlers, so discoverability becomes a number you can put in front of an owner rather than a vibe. Our first live hotel page was found and crawled by Bing and Google within a day of going up; the live directory is at stay.peaqplus.com.
It also automates step 5, weekly instead of monthly: the same panel of traveller questions goes to Perplexity, ChatGPT, Claude and Gemini every week — questions that never contain the hotel’s name, so they measure being found, not being repeated — and the module records whether you were mentioned, which competitors were recommended instead, and whether the link in the answer was yours or an OTA’s. Most hotels’ first measurement is 0%. That’s not a sales problem for us to spin: it’s the baseline nobody had, and the number every following week is measured against.
That’s the disclosure. The five steps above stand on their own, with or without us.
When to ignore this
- If your hotel sells out year-round on repeat guests and word of mouth, this channel adds little. Enviable; carry on.
- If the basics are broken — a booking engine guests abandon, reviews going unanswered — fix those first. Answer engines amplify what exists; they don’t repair it.
- If someone pitches AI discoverability as a replacement for your direct-booking strategy, close the meeting. It’s a compounding side channel, not a new main funnel.
Where to go from here
This channel is growing a next step: assistants that don’t just answer but complete the booking. What that changes — and the seven checks that make a hotel bookable by proxy — is covered in When the Guest Is an AI. And when you’re ready to work the channel as a discipline — the website audit, the content layer, the measurement loop — the practical guide is GEO for Hotels.
The Discovery page shows what a machine-readable hotel page looks like in practice, and the live directory has real examples you can read the way an assistant would.
If you want the broader context of what AI does and doesn’t change in hotel revenue work, What AI Actually Does in Hotel Revenue Management is the companion piece.
Or test it yourself tonight: ask three AI assistants for a hotel recommendation in your city, in your segment. If you appear — check which source got you there. If you don’t, you’ve just met the gap this article is about.
The travelers are already asking. The only question is whose facts the answers are built from.
Reading is one thing — knowing your next step is another. Answer one question and we hand you the guide that matches where your hotel is today. Free, delivered by email.
Find your guide →