We Checked 15 Big Travel Sites. Only Two Gave an AI Agent a Readable Homepage.

In April, Expedia Group published a survey of more than 5,700 travelers in the US, UK and India, run by YouGov. 53% said they are comfortable with AI suggesting travel options. Only 8% said they are comfortable booking through an AI platform.

So for now, the job AI does in travel is research: which hotel, which flight, which neighborhood. To do that job, an assistant has to read the travel site. We wanted to know what it gets when it tries.

How we checked

This is the second post in our weekly series where we run our free agent-readiness check on one industry's best-known sites. Last week was 20 US retailers.

On November 12, 2026 we checked the homepages of 15 airlines, online travel agencies and hotel chains, one site at a time, then ran all 15 a second time. Both runs gave the same result for every site. Sites that failed with a server error or could not be reached got a third try.

The check fetches the homepage the way a simple AI agent would, with a user agent that names us (X1Y1-Tools/1.0). It reads robots.txt rules for the AI search and assistant bots, counts the words in the raw HTML before any JavaScript runs, and looks for structured data, metadata, a sitemap and llms.txt. It scores out of 100.

The same limits as last week apply. Our checker is not ChatGPT, and many bot protection services keep allow lists of verified AI bots, so a site that turned us away may still let OpenAI or Anthropic through. And a homepage is not a search results page. This is a first look, not an audit.

The results

9 of the 15 never got far enough to score. HTTP 520 is a Cloudflare error code that means the site's own server sent back something Cloudflare could not use, so we count it as a failure to read the page, not as a deliberate block.

An earlier, wider pass on November 8 found the same walls at Expedia, Tripadvisor, Marriott, Hyatt and American Airlines, plus Vrbo and IHG.

The doors are open. The pages are empty.

Here is the finding we did not expect. All 6 sites we could read allow every AI assistant bot we check for in robots.txt: OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User. Nobody in this set is saying no on purpose the way Amazon did last week.

But 4 of those 6 had no readable text in their homepage HTML:

A browser runs the page's JavaScript and fills it in. A simple agent that reads HTML gets an empty shell. Booking.com's homepage did not even have a <title>. So for 13 of the 15 sites, an agent like ours ended up with either a block or a blank page.

Airbnb needs a footnote. Our checker received a page titled "Redirecting to www.airbnb.co.uk", a redirect based on location, so its 30 reflects that page, not the US homepage. Its robots.txt is still worth noting: it allows all the assistant bots but blocks GPTBot, ClaudeBot and Applebot-Extended, which are used for AI training. That is a deliberate split, and a reasonable one: let assistants read the site for a user, don't let crawlers take it for training.

What Kayak does right

Kayak scored 90. It lets the assistant bots in, serves a homepage with real text in the HTML, has JSON-LD structured data (WebSite, Organization and FAQPage), a sitemap, a title and description, and it is the only one of the 6 we could read that serves an llms.txt file, at about 7,500 characters.

Priceline scored 75 with readable HTML and TravelAgency structured data. It lost points for having no llms.txt and a sitemap that lists only 13 URLs and is not linked from robots.txt.

None of the 6 offered a markdown version of the page, and none had a Content-Signal line in robots.txt. Those are newer signals and nobody has shown they change bookings. We would fix readable HTML long before either.

Where this is overhyped

A few things we would not claim from this data:

• That Expedia or Marriott "block AI". Our checker got a bot-protection response. Their policy for verified AI bots may be different.
• That Delta or Booking.com are invisible to ChatGPT. Some assistants render JavaScript, and big brands are in training data anyway. Our point is narrower: a plain HTML reader gets nothing.
• That a score of 90 brings bookings. With 8% of travelers comfortable letting AI book, a readable homepage helps with research and recommendations, not checkout, for now.

What a smaller travel business can do with this

This is the part we care about. A boutique hotel, a tour operator or a small airline does not have Booking.com's brand to fall back on. If an assistant cannot read your page, it recommends someone it can read.

What we would check this week:

1. Turn JavaScript off and load your homepage and your main room or tour page. Name, location, price range and what you offer should be in the HTML. If the page is blank, ask your developer about server-side rendering or prerendering.
2. robots.txt. Make sure ChatGPT-User, Claude-User, Perplexity-User and OAI-SearchBot are not disallowed. Decide separately about training crawlers like GPTBot and ClaudeBot, the way Airbnb did.
3. Bot protection rules. If you use Cloudflare or a similar service, check whether verified AI bots are allowed through or get the same challenge as a scraper.
4. Structured data. Hotel, LodgingBusiness or TouristTrip JSON-LD with your address and price range.
5. One test. Ask ChatGPT and Perplexity about your own property by name and see whether they can describe it accurately.

You can run the first three on your own site with our agent-readiness check. It is free and takes about ten seconds.

Next week

Healthcare is next, then banking. If there is a site you want in the set, tell us.

Originally published at x1y1.co.