When a traveller asks an assistant, the hotel should be the answer.
A guest planning a trip increasingly asks an assistant first — where to stay in Hammamet, is there a hotel near the marina with a spa, what time does its restaurant close. The assistant answers from what it can read. Most hotel information is in a booking engine or a JavaScript app, and most assistants do not run JavaScript.
The question
“A hotel near the marina in Hammamet, with a spa?”
What it can read
- Rooms, and what is in them
- The restaurants, and when they open
- The spa, and what it offers
- Where the hotel is
Served as text, not built by JavaScript afterwards.
The answer
The hotel is in the answer, with its hours and where it is — because those were readable.
A diagram, not a screenshot of an assistant.
01AI assistants
Readable without a browser
The hotel’s pages are rendered to plain HTML, so what a guest sees and what an assistant can read are the same thing. The restaurants, the opening hours, the spa treatments and the activities are in the text of the page rather than fetched after it loads — which is the whole difference between being quoted and being skipped.
- Pages served as readable HTML, not assembled in the browser
- Opening hours and services in the text, not behind a request
- Structured data describing what the page actually contains
02AI assistants
Answered from the hotel’s own documents
The concierge answers guests from what the hotel has written — its menus, its policies, its hours — rather than from a general model’s impression of what a hotel is like. When it does not know, it says so and passes the question to the team, which is the behaviour that keeps a wrong answer from becoming a complaint at the desk.