Fundamentals · 11 min read · September 25, 2026 · Hafsa Tariq, gaash.ai
AI visibility for hotels: named for which conditions, linked to whose page
For a hotel, AI visibility is whether an assistant names the property when a traveler stacks conditions (parking, a dog, a family room, a late arrival, a short walk to the venue), whether it says you meet each one, and which page it links under your name. An answer can name you and still send the traveler to a booking platform's listing, so the reading records both.
What a hotel is named for: the conditions in the traveler's question
The questions worth measuring for a hotel carry a place, a kind of trip and a stack of conditions: "hotel near the convention center in [city] with parking", "family room for four with breakfast near [attraction]", "somewhere in [town] that takes a large dog and lets me check in after midnight". An answer can name your property for one of those conditions and stay silent on the rest. It can also name you for a condition you do not meet.
So a hotel is never simply named or not named. It is named for something: the parking, the family room, the late check-in, the walk to the venue. The reading described here keeps that detail. For each answer it asks which of the traveler's conditions the answer said you meet, and whether each of those statements was true on the day of the reading. What your own pages have to state for each condition is set out in the overview for hotels; this page is about finding out whether the answers use it.
An answer can also send the traveler somewhere. One given with web search on can cite or link the pages it drew on, and the link under your name might go to your room page, a booking platform's listing of your property, a review site's page about you or a map listing. A traveler who clicks lands wherever that link goes. Named with your own page linked and named with a platform's listing linked are the same verdict in the score and two different outcomes for the booking, so the record keeps them apart.
When a property counts as named, recommended or first
Every answer can get three verdicts: named, recommended and first. Our published method records two of them, named and first, and a hotel keeping its own record can add the third. Read for a hotel, each one has a boundary worth writing down before the first reading, so that two people recording the same answer reach the same verdict.
| Verdict | Counts for your property when | Does not count |
|---|---|---|
| Named | Your property appears anywhere in the answer, under a name a traveler would recognize as yours, including in a list the answer gives without comment | The platform's name on its own; a sister property in the same group; the restaurant or spa in your building named for dinner or a treatment, which the reading treats as a separate business |
| Recommended | The answer puts you forward for the trip asked about and says you meet at least one of the traveler's conditions | A mention in passing, a mention as a caution ("parking is limited"), or a mention for a condition the answer says you do not meet |
| First | Yours is the first property the reader meets in the answer, whether in a list or in prose | The top of a second list of other options, added after the answer has already recommended someone else |
How a visibility score is built is on the methodology page: the share of the panel's questions in which the property was named, with no weighting, and a hotel changes nothing in that. What a hotel adds is the condition. A property recommended for its parking in every answer and never for its family rooms can post a respectable score while families are sent elsewhere, and only a column for conditions shows it.
The page under your name: your site, a listing, a review site or a map
When an answer names your property and links a page, that page is where the trip continues. Six cases are worth telling apart.
| The answer linked | Where a traveler who clicks lands | What it tells you |
|---|---|---|
| Your room page or policies page | On your own site, on the page that states the condition | The page you wrote for that condition was the one cited |
| Your homepage | On your own site, a click or two away from the answer to the condition | Your site was cited, but not the page that answers the question |
| A booking platform's listing of your property | On the platform, beside its rates and its booking button | The copy of your facts that got cited is the one the platform keeps, not the one you write |
| A review site's page about you | Among guests' reviews of your property | Guests' words were the cited evidence for the condition |
| A map listing | On the listing, with directions, a phone number and a link to your site | The listing's facts are what the traveler reads next, so they have to match your policies page |
| No link | Nowhere; the traveler has to search for your name | There is nothing to check the answer against, and it may have come from the model's training data rather than a page fetched that day |
We have no figure for how often each case happens to hotels. The probe attached to our audit of 9,483 small business websites (one buyer question each for 9,240 businesses across 40 trades, put to Claude with web retrieval on, late August to early September 2026) counted names: 99.4% of the names given were actual businesses and 0.6% were directories or aggregators. It published nothing about which page each answer linked, and it has no hotel cut. The link column is how you get that number for your own property.
A listing in the link column is not a failed reading. You were named, and the verdict counts. What it raises is a question about your own room page, which has to be the better place to land for the travelers who do reach it, and that is the subject of direct bookings from AI answers.
Twelve traveler questions, by trip type, condition and place
The panel below is a template for one property. It has twelve lines, the size of a paying client's panel under our method. Fill the brackets once, with your own city, the venues and sights your guests come for and the conditions you meet. Then leave the wording alone: setting up prompt monitoring explains why a reworded line starts a new series instead of continuing the old one.
| Trip | Conditions | The line, as a traveler might type it | Tests |
|---|---|---|---|
| Business | Walking distance | "hotel within walking distance of [convention center] in [city]" | The shortlist |
| Business | Parking, late arrival | "hotel in [city] with parking where I can check in after 11pm" | The shortlist |
| Family | Family room, breakfast | "family room for [n] with breakfast included near [attraction]" | The shortlist |
| Family | Connecting rooms, pool | "hotels in [city] with connecting rooms and a pool for a family of five" | The shortlist |
| With a dog | Pets, the fee | "dog friendly hotel in [neighborhood] with no pet fee" | The shortlist |
| With a dog | Large dog, ground floor | "hotel near [park or trail] that takes a big dog, ground floor room" | The shortlist |
| On the road | Late arrival, budget | "motel off [highway] near [town] under [amount] with check-in after midnight" | The shortlist |
| Event guest | Distance, shuttle | "hotel near [venue] in [town] with a shuttle for wedding guests" | The shortlist |
| Couple | Quiet | "quiet bed and breakfast near [town] for a weekend away" | The shortlist |
| Access needs | Step-free, roll-in shower | "hotel in [city] with a roll-in shower and a step-free entrance" | The shortlist |
| Any | A policy, by name | "does [your property] allow dogs, and what does it charge" | The facts given about you |
| Any | A comparison | "[your property] or [other property] for a family weekend in [city]" | How you are set against a rival |
Ten of the twelve test whether you make the shortlist for a traveler who does not know your name yet. The last two test what an answer says once it has your name. Weight the lines toward the trips your guests take: a property by the airport and a guest house on a lake need different panels, and a line no guest of yours would type measures nothing you can act on.
Give one of the twelve lines to a condition you do not meet, if it is one your front desk gets asked about. An answer that recommends you for it has stated something false about your property, and it is better to find that in a record than from a guest at the desk who booked on the strength of it.
Three columns a hotel adds to each answer's record
Each answer gets the standard record: named, recommended, first, who else was named, and the whole answer pasted in, in the format the prompt monitoring article sets out. A hotel's answers can get the name right and everything attached to it wrong, so the record carries three more columns.
| Column | What to write | What it catches |
|---|---|---|
| Conditions granted | Each condition in the line; whether the answer said you meet it (yes, no, not stated); whether that is true on your policies page that day | A condition you meet that answers never credit you with, and one you do not meet that they credit anyway |
| Linked or cited | Your room page, your homepage, a platform listing (name the platform), a review site, a map listing, or none | Answers that name you and send the traveler somewhere else to book |
| Policy stated wrongly | Any check-in time, pet rule, parking charge, breakfast hour or cancellation term that differs from your policies page, with both wordings | An old policy still in circulation, or a sister property's rules given as yours |
date: [YYYY-MM-DD]
assistant, run: [ChatGPT | Claude | Gemini | Perplexity], run [1 to 3]
line: hotel in [city] with parking where I can check in after 11pm
named: [yes | no]
recommended: [yes | no]
first: [yes | no]
conditions: parking: said [yes | no | not stated], true [yes | no]
check-in after 11pm: said [yes | no | not stated], true [yes | no]
linked or cited: [room page | homepage | platform listing: [platform] | review site: [site] | map listing | none]
policy stated: answer said "[words]"; policies page says "[words]"; [matches | differs]
others named: [property A], [property B]
answer: (pasted in full, unedited)Check the conditions column against your policies page as it reads on the day of the reading, not against memory. If the pet fee changed last month, an answer quoting the old fee has been wrong since the day it changed, and the record should say so.
What a month of these records tells a hotelier
Read the columns before the score. A condition you meet that no answer credits you with points at the page that should state it: the parking section, the pet policy, the page for the family room. A condition answers credit you with that you do not meet is a wrong fact, and so is every entry in the policy column. Both are traced the way the article on correcting what AI gets wrong describes, starting with your own pages and then the listings that repeat them.
The link column gives you a second count: of the answers that named you, how many linked your own site and how many linked a listing. The score leaves it out, because the published score counts whether the property was named and nothing else. If you care where the booking happens, it is the column to watch.
The first month is a baseline, not a verdict on the hotel. The only large reading we have published is the probe's: Claude named 14.9% of 9,240 businesses on one buyer question each. That is a Claude figure from one question on one day, across all 40 trades with no hotel cut. It shows a direction, not a rate, and it gives no grounds for reading a low first month as a failure.
Comparing one month with the next, and with the same month a year on, has its own complications for a hotel, and measuring AI mentions for hotels deals with them. Whatever the comparison, the record above is what it is made from.
Common questions
Should some of the panel's questions name our property?
Two or three, no more. Lines such as "does [your property] allow dogs" or "what time is check-in at [your property]" test the facts an answer gives about you, and a wrong policy is easiest to spot in them. They say nothing about the shortlist, because the traveler supplied your name, so read them apart from the other lines rather than letting them lift your mention rate.
An answer named us but linked a booking platform. Is that a bad result?
It is a result, and it counts in the score: you were named. The link column records where a traveler who clicked would have landed, which is a separate question. If the answers that name you keep linking a listing, compare that listing with your own room page and policies page, fact by fact, and read the article on direct bookings for what your page needs when a traveler lands on it.
We run three small properties in one town. One panel or three?
Three, one per property, because each is asked about for its own street, its own rooms and its own rules. Add a note in the policy column whenever an answer gives one property's check-in time, pet rule or phone number for another. A mix-up between sister properties is a wrong fact like any other, and it is traced from the pages and listings that describe each property.
Should some lines be in the languages our guests speak?
If a real share of your guests plan the trip in another language, add a line or two in it, written by someone who speaks it, and read them as a separate series. The answers may cite different pages, such as a translated page if you have one. We have published no measurement of how answers differ by language, so compare those lines only with themselves.
Sources
The probe figures are this site's own, from one question per business put to Claude once, across all 40 trades; there is no hotel cut, and no figure for which pages the answers linked. The score is the published method's, which records named and first and does not weight them; the recommended verdict and the hotel's columns are additions for your own record.
Sources: our audit of 9,483 small-business sites and its Claude probe, with the method · gaash.ai methodology: the panel, the assistant and the score
Read on