When a guest asks an AI assistant to find them a place to stay, the assistant does roughly what the guest would do: it visits websites, compares options, and checks prices. Except it does all of it in seconds, and it does not experience your website at all. It reads it. Whether your site is readable, in the specific way machines read, increasingly decides whether you get recommended or silently skipped.
This article is about that reader: what it sees, what it ignores, and what it needs from you.
An AI assistant does not experience your website; it reads it. Many AI crawlers and assistants read the raw response your server returns and never run the page's JavaScript. So if your prices, availability, and property cards are fetched into the page afterward, in the guest's browser, the machine's copy of your site has holes exactly where the money is. It judges you on what arrived: the words, the numbers, and the structured data underneath them. The thirty-second test is to view the page source on a listing page and search for a nightly rate. If the number is not in the source, the machine never saw it.
What can a machine read on your listing page?
Paste a property page or a search results page. We read it the way an AI crawler does, without running its JavaScript, and check what it gives a machine to work with: eight photos, a real rate, precise coordinates, and on a results page, whether every row carries a price an assistant can compare.
Your second customer
Your website was designed for one kind of visitor. It now has two, and they could hardly be more different.
A human guest scrolls, looks at photos, forms an impression, and forgives a lot along the way: a spinner here, a slow gallery there. An AI assistant fetches your page, reads what came back, and moves on. It does not wait for anything. It does not see your design. It judges you entirely on the substance of what the server sent: the words, the numbers, and the structured data underneath them.
The differences that matter
- 01
A guest waits for the page to finish. A machine reads what arrived. If parts of your page load in late, the machine has already read a page without them.
- 02
A guest sees the vibe. A machine reads the claims. Photography and brand win humans; machines weigh verifiable specifics: a nightly rate with dates attached, a real availability window, a cancellation policy it can quote.
- 03
A guest might come back. A machine recommends someone else. There is no second impression at fetch time. An assistant building a shortlist reads your page once, and if the answer was not there, you are not on the list.
Machines read the page the server sends
Here is the single most important mechanical fact: many AI crawlers and assistants read the raw response your server returns and never run the page’s JavaScript.
That matters because of how most booking sites are built. The server sends a page frame, and then scripts running in the guest’s browser fetch the real substance (prices, availability, property cards) into it afterward. A human never notices, because the placeholders fill in while they look at the hero photo. A machine reads the frame. Its copy of your site has holes exactly where the money is.
The test takes thirty seconds: open a property page, view the page source, the raw HTML rather than the rendered page, and search for a nightly rate. If the number is not in the source, the machine never saw it.
You can go one step further and paste the page into Google’s Rich Results Test, which shows whether your listings carry the structured data machines use to render them as product cards.
Machines want claims they can act on
A human reads “luxury oceanfront escape” and feels something. A machine reads it and extracts nothing it can use. What a machine can use is a dated offer, a real availability window, a total price, a policy: claims specific enough to verify and act on. And “act on” is the operative phrase: assistants increasingly do more than recommend. They compare your rate against an OTA’s, present the difference to a guest, and carry the booking forward. Every claim on the page needs to be one you would stand behind at checkout, because the machine will treat it that way.
Making your claims machine-verifiable is a discipline of its own. That is agentic engine optimization, and for assistants that query live systems directly rather than reading pages, it extends to running an MCP server. This article’s point is narrower: none of that stack matters if the page the machine reads is empty where the facts should be.
The good news: both customers want the same thing
Nothing machines need is at odds with what guests need. A page that arrives finished, price included and no spinner, converts humans better and is the same page a machine can read. Structured data that renders your properties as product cards for AI assistants is the same markup that earns rich results in classic search. Serving everything from your own domain protects both your brand equity and your citations.
The work is not “optimize for AI.” The work is building your site the way modern commerce builds sites: assembled on the server, specific in its claims, current in its numbers. Whether your current platform can do that, and what to do if it cannot, is the subject of our companion article: WordPress or a modern web framework?
Frequently asked questions
Do AI assistants really visit my website?
Yes, two ways: crawlers gather pages ahead of time, and assistants fetch pages live while answering a guest. Both read what the server returns. If you check your logs, you will likely find them already there.
Will better photos help me get recommended?
Photos still win the human once the recommendation is made, but the machine deciding whether to recommend you is reading text and data. You need both, doing different jobs.
Is this just SEO with a new name?
No. SEO earns you a position on a results page a human then reads. A machine reader skips the results page: it evaluates your actual pages and acts on what it verifies. The overlap is real (crawlability, structured data), but the bar is higher, because the machine needs answers, not just relevance.
How do I find out what machines see on my site?
View the page source on a property page and look for your prices, then run the page through Google’s Rich Results Test. If the rate is not in the source and the test shows no product data, machines are reading an empty shell.
Where do you stand today?
Run the thirty-second test on your own listing page. If the facts are in the source, you are ahead of most of the industry. If they are not, your site has a second customer it has never once served.