How agent-ready are the leading travel (hotel companies) brands?

Where the leaders fall short.

Median score per area, across the brands where the area could be measured.

The most common gaps.

1. 1. WebMCP tools an agent can call 10 of 10 sites fail this check · Act An agent arriving on this page is given no actions at all. It has to read the screen and guess where to click, which is slow and frequently wrong.Fix: Register WebMCP tools for the two or three things a customer actually comes here to do: check availability, request a quote, book, add to cart, apply a promotion. Each tool is a name, a short description and a typed input. The browser session already carries the login, so there are no API keys or tokens to issue.How to fix it
2. 2. Machine readable capability manifest 10 of 10 sites fail this check · Act There is no manifest telling an agent what your site is capable of.Fix: Publish /.well-known/mcp.json describing your available tools. It is a small static file and it is how an agent decides you are worth interacting with rather than scraping.How to fix it
3. 3. Rules for agents are written down 10 of 10 sites fail this check · Delegate You have not written down what agents may do on your site. A careful personal agent, unsure whether it may book or cancel for someone, asks its person instead of doing the job.Fix: Add a "## Agents" section to /llms.txt saying what an agent may do for a customer (book, reschedule, cancel), what needs the customer to confirm, and where to send them when it cannot. Add a Content-Signal line to robots.txt (for example: Content-Signal: search=yes, ai-input=yes, ai-train=no) so your use terms are machine readable.How to fix it
4. 4. llms.txt briefing for AI 9 of 10 sites fail this check · Comprehend
5. 5. A customer can give their agent access 6 of 7 sites fail this check · Delegate
6. 6. No captcha blocking the last step 7 of 9 sites fail this check · Complete
7. 7. Machine readable business facts 6 of 10 sites fail this check · Comprehend
8. 8. Forms a machine can fill in 6 of 10 sites fail this check · Act
9. 9. Account jobs an agent can do 4 of 7 sites fail this check · Delegate
10. 10. Site serves a request from an agent 12 of 22 sites fail this check · Reach

Every score, no names.

Each bar is one of the 22 brands we could scan, highest first. Brands are counted, not named. 3 more could not be scanned at all.

How we chose and measured.

Who: Ranks hotel companies by 2025 year-end room count; this list keeps only U.S.- and Canada-headquartered companies that sell rooms to travellers under their own brand (third-party management firms skipped), so source ranks run to 159. Source: HOTELS magazine, HOTELS' 325: The List 2026 (2026-07). Sneeze It clients are left out.

When: each brand's main website was scanned once, October 25, 2026, by AgentReadyBot, which identifies itself on every request, obeys robots.txt and never tries to get past a bot wall. Public pages only. Sites change, so a brand's score today may differ. We rescan every month. The full method is on the methodology page.