Overview
A scan answers three questions about one hotel's website: how ready it is for AI agents, what is stopping them, and what to fix first. It measures the site against What Travel Agent Readiness asks of a hotel's website. The result is a score out of 100, split across three stages, with a ranked list of fixes that names who makes each one.
1 · The scan
What a scan does
A scan meets a hotel the way agents do. It finds the hotel's pages from the homepage, the links on it and the sitemap. It opens each page the ways agents open them: a quick fetch that does not run scripts, which is how many AI assistants read the web, and a full browser. It follows the booking link as a guest would, with sample dates a few months ahead for two adults, and moves the dates later if nothing is available. It reads the pages for the twelve stay facts a traveller asks about, from check-in times to parking.
AI models help in two places: walking the booking flow and reading pages for stay facts. Fixed rules then turn what the scan recorded into points. The scan keeps every page and reading a score relies on, so each point can be traced to its evidence and the score worked out again from the same record.
2 · Stages
Three stages, 100 points
| Stage | The question | Points |
|---|---|---|
| Find | Can an agent find the hotel? | 30 |
| Read | Can it read the facts of a stay? | 50 |
| Book | Can it book a room? | 20 |
Find and Read measure the hotel's own website, 80 points between them. Book measures the booking engine. Read carries half the score, because it is where an agent gets its answer or does not.
Read's content checks look at the pages as a browser shows them, so a site that turns agents away still earns Read points for what its pages say. The block itself is scored in Find, where it also holds the whole score down.
A hotel with no online booking earns nothing in Book, and those 20 points do not move to Find or Read, so its score can reach at most 80.
3 · Points
How points are earned
Each requirement that counts has a fixed share of its stage's points, set by its level in the standard. Within each stage, Required requirements (MUST in the standard) carry the most, Recommended ones (SHOULD) less, and Optional ones (MAY) the least. A requirement is judged by a few checks, and each check has one of five outcomes.
| Outcome | What it means for the points |
|---|---|
| Met | Full credit. |
| Partly met | A stated share, such as nine of the twelve stay facts answered. |
| Not met | No credit. |
| Doesn't apply | Left out, so it neither helps nor hurts, such as booking checks for a hotel with no online booking. |
| Couldn't tell | No credit, and flagged. It never holds the score down, so a site is not punished for what the scan could not see. |
A requirement earns its full share only when every check is met, and any shortfall costs at least one point. The score is the sum of what every requirement earned.
4 · Ceilings
When a score is held
Required requirements describe the failures that stop an agent outright or make a failure look like success: turning agents away, hiding the facts behind scripts, or answering with misleading status codes and disguised bot checks. When one of them fails, losing its points is not enough, so the score is held at a ceiling until it passes.
| Held at | When |
|---|---|
| 27 | A Required check fails on the hotel's own website. |
| 80 | A Required check fails only on the booking engine's site. |
The lowest ceiling applies. Where a failure could be a passing glitch, the scan fetches the page again and holds the score only if the failure repeats. Partly met and couldn't-tell outcomes never hold a score.
The standard calls a site Visible when every Required requirement in Find and Read is met. A held score therefore means the site is not yet Visible. The reverse does not follow: a Required check that was only partly met, or that the scan could not settle, does not hold the score but still leaves the site short of Visible.
5 · Owners
Who fixes each finding
Every check that falls short becomes a finding, and each finding names who can fix it.
| Owner | When | What the fix looks like |
|---|---|---|
| The hotel | A Content requirement falls short on the hotel's own pages. | Change what the site says. |
| The website | A Technical requirement falls short on the hotel's own site. | Change how the site is built or served. |
| The booking-engine vendor | A check falls short on the booking engine's site. | The vendor changes its engine. |
This follows the standard's two categories: Content requirements fall to the hotel and Technical ones to the website, unless the problem is on the booking engine's site.
Findings are listed by what fixing them is worth. Those holding the score down come first, lowest ceiling first, then the rest by the points each fix would add on its own.
6 · Not counted
What does not count yet
Four requirements describe agent interfaces, ways for an agent to read or book without using the web pages. They are optional in the standard and earn no points yet, because their benefit to agents has not been measured sufficiently. Deferred requirements cover steps a scan cannot observe and are not assessed.
7 · Versions
Keeping scores comparable
Every score names the version of the rules it was worked out under. When the rules change, a score can be worked out again from the evidence already recorded, without visiting the site again, and the Index lists only hotels scored under the current version.
