Ask ChatGPT where to stay in a US city and it answers with a ranked list of about nine hotels. Not a single pick, and not ten blue links. That is the good news: a list has positions, positions can be measured, and measured positions can be improved one at a time.
Those numbers come from the AI Hotel Ranking, our public census — 2,322 of the questions travelers actually ask, put to ChatGPT across 18 US markets, with every hotel it names recorded. Scarcity is real; it is just not three slots, and nobody owns the dozen that matter. So the useful question is not "am I in the answer". It is how often, and how high.
The playbook
1. Read the board for your market
The ranking is public and needs no account. Find your comp set, find yourself, and read the rank — not just whether you are listed. A hotel named in six answers out of ten but averaging ninth has a different problem from one that is never named at all, and the two do not share a fix.
2. Scan your own property
A market board is not your property. Your surface is the same questions asked as a couple, a family, a business traveler and a budget-conscious guest, across the assistants travelers actually use. Typing your own name into ChatGPT is one question, one engine, one roll of the dice. A free scan asks the questions travelers use for your property and shows where you are missing. It asks for a work email, nothing else.
3. Work outward from what you control
Your own site first. Then your structured data — the machine-readable facts an engine can verify instead of guess: rates, amenities, policies, location. Then OTA listings, reviews, local and maps, and editorial coverage last. The order is not arbitrary: you can change your own page this afternoon; you cannot change a press mention at all.
4. Trace every job back to a finding
A job worth doing points at a specific observation — which questions, which engines, what came back — not a hunch that the page could be better. AI Visibility turns each finding into one ranked action: where the work lands, what it touches, and one to six concrete steps out in the real world. It is in the dashboard on every paid plan.
5. Hand each job to the assistant you already use
You do not need a new tool to do the writing. Give ChatGPT, Claude or Gemini the finding, the steps and your property URL, and ask for a plan before it writes anything. Rewriting a section of your own site is a job assistants are genuinely good at; structured data is better still, because markup is exact. But none of them can log into a Booking.com or Expedia extranet — there, the win is that it writes the exact text for every field and you paste it. Review replies and press pitches too: drafted in full, sent by you. Our working rules for this are public.
An assistant asked for a capacity, a distance or a rate it cannot find will produce a plausible number. A round number is still an invention, and it is your front desk that is accountable for it at check-in. Tell it to leave a marked placeholder rather than guess, and review everything before it publishes.
6. Re-measure, and give it time
The verdict on whether an edit worked comes from the next measurement, not from ticking the job off. One caution: ask an AI the same question twice and you can get different answers back, so reading your numbers the morning after a change means reading noise. Give it a few scans — and expect movement in position, not a guarantee of one.
One move that stays last
Being named and then handed to an OTA is half the job done for you and all of it for someone else. Once the visibility work is holding, the thing left to close is the booking path: an assistant that recommends you can also book you, the booking lands in your hotel's own system, and the guest relationship stays yours. That is what AI Direct Bookings is for — after the list, not instead of it.
AI recommends what it can verify — about nine hotels at a time, in order. The work is not getting into the answer. It is climbing the list, one measured fix at a time.