What travelers ask an assistant before they book a hotel

The questions a traveler types before booking

The booking starts as a question with constraints in it, not a hotel name. Typed the way people type: "quiet hotel near the convention center in Denver with parking", "small hotel in Portland that takes dogs and includes breakfast", "hotels within walking distance of the Vancouver cruise terminal with late check-in". Then, once a name is on the table, a second kind: "does this hotel have free parking", "can I check in at midnight", "is breakfast included in the rate".

An AI assistant answering the first kind builds a shortlist. It searches, fetches a handful of pages, and names the hotels whose pages match the constraints it was given. Answering the second kind, it fetches pages about one hotel and reads for the fact. In both cases the material is text it can retrieve at the moment of asking, and your own site is one candidate among the booking platforms, the review sites and the map listing.

The constraints are the point. "Quiet" is a claim every hotel makes about itself, so what makes it checkable is evidence: which rooms face the courtyard, whether there is a bar under the guest floors, what guests wrote. "Parking" is not a yes or a no; it is on-site or off, included or charged, valet or self-park, with a height limit or without. A site that states the specifics can be quoted. A site that says "convenient parking available" gives an answer nothing to check.

What your hotel's pages have to give an answer

The facts the traveler asked about, stated somewhere the assistant can read them, in a form it can attribute to you. That takes three things: the facts have to exist as text, the text has to sit on a page a crawler receives, and the identity on that page has to agree with the identity everywhere else.

The first part fails most often. Hotel sites are written to sell a feeling, and the facts that decide a booking (the check-in window, what happens after the desk closes, the pet policy, the breakfast hours, the distance to the thing the traveler is in town for) end up scattered across a booking widget, a PDF fact sheet and a reservations phone line. If your site does not state your check-in time, an answer can only take it from a platform listing, which may be out of date, or leave you off a list built around "late check-in".

The third part is identity. If your homepage uses one form of the name, the contact page adds "& Suites", and the map listing gives a third phone number, anyone reading all three has to work out whether they describe one hotel, and an answer may treat them as two. The fix is one name, one address, one main number, written identically on every page and in every listing you control. Google's Business Profile guidelines ask for the real name and the real address, which makes the profile a good place to start the comparison.

Reviews are the evidence the assistant uses for the adjectives. "Quiet" on your homepage is marketing. "The courtyard rooms were silent even on a Saturday", in a guest's words, quoted on your page with a first name and a month, is a fact about the hotel that someone other than you stated. A page with no review text and no rating markup asks the assistant to take your word for everything, and there is no reason it should.

What the hotel's site has to state, and where

Each row below is a traveler's question, the fact that answers it, and where that fact belongs on your site: a sentence or two of plain text on a page a crawler receives, not a line inside a booking widget. The same facts go into LocalBusiness schema using the LodgingBusiness type or its Hotel subtype from schema.org's LocalBusiness tree, which has properties for check-in and check-out times, pets, amenities, address, phone and geo. Google's local business documentation names name, address, telephone, opening hours, geo and price range as the properties it reads. Write the visible text first and let the markup repeat it.

None of this needs a new website. It needs one page, usually called something like "Good to know" or "Before you arrive", where each question is a heading and the answer sits under it in two or three sentences. That is an answer page, and it is the page on your site an assistant can quote most directly.

What our crawl found, and how it applies to hotels

In our audit of 9,483 small business websites across 40 trades and 90 metros in the US, Canada, Europe and the UAE, run between late August and early September 2026, 61.5% had no question-and-answer content of any kind: no FAQ section, no FAQPage markup, nothing shaped like the thing a buyer asks. Hotels were among the trades. We do not publish a hotel-only cut of that figure, so read it as the state of small business sites generally.

43.1% were inconsistent about who and where they are across their own homepage, contact page and about page, and 36.9% published more than one phone number across those three pages. For a hotel with a reservations line, a front desk line and a restaurant line, that happens by accident. The fix is to label the numbers rather than remove them: one marked as the hotel's main number, and that one repeated in the markup.

64.7% showed no review signal a crawler could read, meaning no rating markup and no review quoted in the page text, and only 8.9% carried rating schema. Most of those businesses have reviews on the platforms. The assistant reading their own site does not know that, because the site does not say so. The citability of a hotel page comes down to these three things together: an answer to the question, an identity that holds, and evidence that is not your own.

What to fix first, in order

1. Settle the identity. Pick the one trading name, the one address format and the one main phone number, and write them identically in the footer of every page, on the contact page, in the markup and on the map listing. Label the other numbers for what they are.
2. Write the facts page. Check-in and check-out times, the after-hours procedure, parking, breakfast, pets, accessibility, and the walking minutes to the places guests come for. One question per heading, the answer in plain text under it, no widget, no PDF.
3. Put the same facts into LodgingBusiness or Hotel markup as JSON-LD: name, address, telephone, geo, checkinTime, checkoutTime, petsAllowed, amenityFeature, priceRange. Google's introduction to structured data says the markup has to describe what a visitor can see, so this step comes after the page, never instead of it.
4. Quote guests on the page. Three or four short reviews in the guest's own words, with a first name and a month, chosen for the facts they state (the quiet, the parking, the breakfast), and the platform each one came from.
5. Read your platform listings and your map listing against the facts page and make them agree. Where they cannot, because a platform truncates your pet policy, make sure your own page is the fuller one, since that is the one an assistant can quote in full.
6. Check what a crawler receives. Open your facts page with scripts off and confirm the text is in the HTML, then run the crawler access check against your robots.txt. 7.2% of the sites in our audit served a challenge page instead of the site, and a booking engine's security layer can do that without anyone deciding it.

What not to do

Do not write reviews of your own hotel, and do not copy a booking site's score into your markup. Google's review snippet documentation rules out fake reviews and ratings aggregated from other websites, and three glowing paragraphs that appear on your page and nowhere else give a reader nothing to check them against.

Do not stuff the page with place names. "Hotel near Denver convention center, downtown Denver hotel, Denver hotel with parking" is not a fact about your hotel; it is a list of phrases, and it reads as one. Google does not read the keywords meta tag at all, as its list of supported meta tags shows, and an assistant reading prose wants the sentence that answers the question, not the phrase that matches it.

Do not write a separate version of your site for AI. Google's page on AI features names no extra requirement, file or markup for its AI results, and the assistants that fetch at answer time read the same page a traveler does. The page that helps the traveler decide is the page that gets you named.

And do not expect a promise. No agency, this one included, has a lever inside ChatGPT, Claude, Gemini or Perplexity. What can be done is to measure how often your hotel is named on a fixed set of traveler questions, fix what the measurement shows, and measure again. How we measure it is written down in full.