Answer · Process
How can a hotel or tour operator get recommended by AI assistants?
The answer
A hotel or tour operator gets recommended by AI assistants by publishing, on its own domain, plain answers to the questions a booking site cannot answer about that specific property or trip. Booking sites take the generic searches and will keep taking them. An assistant asked which property suits a particular traveller, or what a trip involves, needs sentences it can lift: who the stay is for, what is included, what the first day looks like, how to book directly. Those sentences have to sit in ordinary HTML on addresses that do not change, and the AI crawlers have to be allowed in. Descriptions on the operator's site, its listings and the guides travellers already use should match. We run World Travel Guide on the same rules, and the property is open for anyone to check.
Why can you not win the generic search?
Because the booking sites and aggregators have more budget behind a single city-plus-hotels query than most operators will spend in a year.
A search for hotels in a named city returns pages built by companies whose whole business is that result. Competing for it is a way to spend money proving the point. The same is true of the broad destination questions, which is a problem we know from the inside: Occam owns and operates World Travel Guide, and the guide long ago stopped chasing the results the booking sites own.
What the booking sites cannot do is answer the second question, the one about your property or your trip specifically. Only you know those answers, and they are what an assistant needs when a traveller asks which option actually fits.
What questions should the operator's own pages answer?
The ones a traveller asks after the booking site has shown them a price.
- Who is the stay or the trip for, and who will not enjoy it? Families with young children, solo travellers, people who want silence, people who want a bar.
- What is included, and what is not? Transfers, meals, guides, equipment, the things that turn up as surprises on the invoice.
- What does the first day look like, hour by hour if that helps?
- How does booking directly work, and what changes if you do?
- What is the cancellation position in plain words?
Each answer should sit under its own heading, with the answer in the first sentence, so a lifted passage still makes sense on its own. Softened marketing copy fails this test. A sentence a manager would say to a guest at the desk passes it.
What has to be true technically?
The sentences have to arrive in the first HTML response, on addresses that stay put, on a site the AI crawlers are allowed to read.
Many hospitality sites are a booking engine with almost no readable copy, or a brochure whose content loads on a click. Neither gives an assistant anything to lift. The fix is editorial and structural at once: put the answers in plain text, keep one permanent address per property or trip, and check the robots file admits the crawlers rather than blocking them by default.
Where the current platform cannot do that, a rebuild comes before any visibility spend, and we say so early. The rules we apply are the ones we run on our own property, described in Booking sites took the search results. The guide stayed worth quoting.
How do you check whether assistants read the site?
Read the server log, then run the test yourself.
Client-side analytics cannot see a crawler. The server log records which crawler fetched which page and when, and that record is the first evidence that the work is reaching the assistants. After that, the test is open to anyone: read a page without JavaScript, then ask an assistant a specific question about the property and see whether the answer draws on the operator's own words or on an aggregator's.
Measurement in this sector also has to span the property system and the brand site, or the partners will keep arguing from different numbers. The sector page for travel and hospitality sets out what we change and what we will not take on.