I want to show you something that took me embarrassingly long to internalize, even after years of doing this for hotels.
The way people search broke. Quietly, and then all at once.
For two decades we optimized for keywords that read like a caveman ordering room service. “Orlando boutique hotel.” “Hotel near convention center.” “Pet friendly hotel downtown.” Two or three words, no punctuation, no soul. We built pages to match those stubs because that is genuinely how people typed into a search box. Short, clipped, telegraphic.
Then people started talking to machines instead of querying them. And they talk the way they actually think.
People stopped typing keywords and started asking questions
Go watch a real person use ChatGPT, Gemini, or the AI panel in Google to plan a trip. They do not type “boutique hotel Orlando.” They type, or say out loud to their phone:
“I’m going to Orlando in March for a wedding, I don’t want to stay near the theme parks, somewhere quiet and walkable with a good bar and free parking, ideally under 200 a night. Any boutique places?”
That is one query. Read it again. It has a location, a date, a vibe, three hard requirements, a budget, and a property-type preference, all in one breath. No human ever typed that into Google in 2015. They would have done five separate searches and pieced it together themselves. Now the machine does the piecing, and the traveler just describes their actual situation like they are talking to a smart friend.
This is the whole shift in a sentence. The unit of search used to be the keyword. The unit of search is now the question. And not a tidy question either, a messy, over-specified, run-on human question.
If your hotel pages are still written to match caveman keywords, you are invisible to the part of the conversation where the decision actually gets made. I dug into exactly that problem in why your hotel might be invisible to ChatGPT, and this post is the practical fix.
Why this matters more for AEO than old-school SEO
Quick honesty check on the numbers, because I refuse to throw fake stats at you. The category names people use for this work pull real US search volume: AEO sits around 27,100 searches a month, AI SEO around 8,100, generative engine optimization around 5,400. “Hotel SEO” itself is a comparatively tiny ~590. The point is not those numbers, it is what they imply: a lot of people, including hoteliers, are scrambling to figure out AI search, and almost nobody has rewritten their actual page copy for it yet.
Here is the mechanical reason conversational queries matter so much for AEO and GEO specifically. When an AI assistant answers a travel question, it is not handing back ten blue links. It is composing a paragraph, and it wants to lift clean, self-contained sentences from somewhere to build that paragraph. If your page literally contains a sentence that answers “is there a quiet boutique hotel near downtown with free parking,” you have handed the model a quotable answer on a plate. If your page just says “Boutique Hotel | Downtown | Free Parking” in a header, you have handed it nothing it can repeat.
Old SEO rewarded pages that matched a keyword. AI search rewards pages that answer a question in a sentence the model can lift and repeat verbatim. Those are not the same craft, and writing for the second one is where almost every hotel site is currently leaving money on the table.
How to actually find the questions guests are asking
You cannot write answers to questions you have not collected. So before any rewriting, I build a question list. This is the unglamorous part and it is the part that separates real work from guessing.
Here is where I pull questions from, in rough order of value:
- Your inbox and front desk. The single best source on earth. Every pre-arrival email, every “quick question before I book” message, every thing the front desk gets asked twice a day. “Can we check in early?” “Is the rooftop open in winter?” “How far is it to walk to the convention center?” These are gold because they are real human phrasings, already in your guests’ words.
- Your reviews and your competitors’ reviews. People tell the truth in reviews. They write “we picked this place because it was the only quiet one within walking distance of the old town” and you just learned exactly how a guest frames their decision.
- The AI tools themselves. Ask ChatGPT and Gemini the kinds of questions a traveler in your market would ask. Watch what they answer, watch which properties get named, watch the follow-up questions they suggest. That suggestion list is a literal map of the conversation.
- Google’s “People Also Ask” and autocomplete. Still useful. Type the start of a question about your area and harvest the completions.
I dump all of this into one big messy list, raw phrasings intact. Do not clean them up into keywords. The whole point is to keep the natural language. “somewhere quiet and walkable” is the asset, not “quiet walkable hotel.”
Restructuring the page: answer first, then elaborate
Once I have the questions, I rewrite the page so it answers them in a structure that both a human skimming on their phone and a machine parsing for a quote can use. The pattern I use over and over is answer-first.
Most hotel copy buries the answer. It opens with three sentences of mood lighting before it tells you anything. “Nestled in the heart of the historic district, our intimate retreat invites you to…” Stop. A traveler asking a pointed question, and the AI trying to serve them, both want the answer in the first sentence.
So instead of a header that says “Location,” I write a sub-heading that is the actual question, and then I answer it immediately:
How far is the hotel from downtown and the convention center?
We are a seven-minute walk from the main square and about twelve minutes on foot to the convention center, with a flat, well-lit route the whole way. If you would rather not walk, a rideshare runs three to four minutes.
See what that does? The sub-heading is phrased like a real question. The first sentence is the answer, with specifics. A model can lift that whole paragraph and it stands on its own with zero surrounding context. A skimming human gets their answer in two seconds. Nobody had to wade through mood lighting.
Compare the two approaches side by side, because the difference is the entire game:
| Old keyword style | Conversational query style |
|---|---|
| Header: “Location & Parking” | Heading: “Is there free parking and how far is downtown?" |
| "Conveniently located downtown with parking available." | "Yes, parking is free for all guests in our private lot, and you are a seven-minute walk from the main square.” |
| Optimized for: downtown hotel parking | Optimized for: the real question a guest asks |
| AI tool can lift: nothing usable | AI tool can lift: a complete, quotable answer |
You keep the vibe writing. Boutique hotels live and die on atmosphere, and I would never strip that out. But the atmosphere becomes the second paragraph, after the answer, not the wall you make people climb before they get information.
Your FAQ section is the highest-leverage real estate on the site
If you do nothing else after reading this, rebuild your FAQ. It is the most natural home for conversational answers because the format already is a question and an answer. The problem is most hotel FAQs are written lazily, with keyword-stuffed questions and one-line non-answers.
Rules I follow for an FAQ that AI tools will actually quote:
- Write each question in the guest’s real voice. Not “Parking availability” but “Do you have free parking, and is it on-site?” The match between their phrasing and yours is what gets you pulled into the answer.
- Make every answer self-contained. Assume it will be read with zero context, ripped out and pasted into an AI response. So the answer repeats enough of the question to stand alone: “Yes, on-site parking is free for all registered guests…” not just “Yes, it’s free.”
- One idea per answer, two to four sentences. Long enough to be specific, short enough to be liftable. If an answer is sprawling into a paragraph of caveats, split it into two questions.
- Be specific and be honest. “Check-in is at 3pm, but if your room is ready earlier we will always let you in, and we hold bags free if it is not” beats a vague “early check-in subject to availability” every time, for both humans and machines.
A boutique property I am thinking of, purely as an illustration, had a single FAQ entry that said “Pets: Yes.” I rewrote it into “Are dogs allowed, and is there a fee or a weight limit?” with a real three-sentence answer about the flat 25 dollar fee, no weight limit, and the two ground-floor rooms nearest the courtyard. That one rewrite turned a dead line into something an AI assistant could confidently use to recommend the place to someone searching for dog-friendly options. I am not going to pretend a single FAQ edit is a revenue miracle, it is one brick. But the wall is made of bricks.
Why this is also your best shot at winning back direct bookings
Here is the strategic part, and it ties straight into the thing every independent hotelier I talk to actually cares about: the OTA tax. You are handing roughly 15 to 25 percent of those booking values to the OTAs in commission, and you would very much like to claw some of that margin back. I broke the actual arithmetic of that down in the book-direct math post, and it is uglier than most owners want to admit.
You are never going to fully escape the OTAs, and anyone who tells you otherwise is selling you something. They are too big, too good at distribution, and frankly they will keep sending you guests. The realistic goal is a healthier mix, more direct bookings, less dependence, more margin you keep. I wrote about how the OTAs out-rank you even for your own situations in how OTAs steal your search.
But here is the opening conversational AI search gives you. When a traveler asks an AI assistant a hyper-specific question, “quiet boutique hotel, walkable, free parking, good bar, under 200,” the OTA’s generic category page is a weak answer. The OTA does not have a self-contained sentence answering that exact combination. Your page can. You can out-specific them on the long, weird, human questions in a way you can never out-muscle them on the broad ones. The detailed conversational query is the seam where a small property can actually get named directly, send the traveler to your own site, and book them without the commission haircut. Turning that visibility into an actual completed booking is its own discipline, which is the whole point of getting your direct booking flow right.
A simple order of operations to start this week
If you want to act on this without boiling the ocean:
- Collect 20 to 30 real questions from your inbox, front desk, and reviews. Keep the exact phrasings.
- Pick your five highest-stakes pages (homepage, rooms, location, your single best room type, the booking page) and add answer-first question sub-headings to each.
- Rebuild the FAQ with real-voice questions and self-contained answers, two to four sentences each.
- Test it. Ask ChatGPT and Gemini the questions you just answered and see whether your property comes up, and whether what they say about you is accurate.
- Repeat monthly. The question list is never finished, because guests keep asking new things.
None of this is a guarantee of a top ranking or instant AI mentions. Nobody can promise you that, and I am suspicious of anyone who does. What rewriting for conversational queries does is dramatically improve your odds of being the page a model can quote, on exactly the high-intent questions where a boutique hotel can win. Pickup can happen in a few weeks, durable visibility usually takes a few months of consistent work plus the reputation signals that make AI tools trust you in the first place.
If you want a hand turning your guests’ real questions into pages and answers that AI search can actually use, that is the core of what we do at our AEO and GEO service, and you can grab a free intro call any time over at the booking page. Bring your weirdest guest questions. Those are the ones worth the most.