What does it mean for a website to be 'readable' by an AI agent?

That a program which doesn't run JavaScript, doesn't interpret your design and can't solve captchas can still extract the data it needs: what you offer, what it costs, which policies apply and how to book. If content only exists after a browser renders the page, for many AI crawlers it doesn't exist.

What is AXO (Agent Experience Optimization)?

A term coined by SiteMinder in 2025 for the discipline of optimising a website for autonomous agents, not just people: complete metadata, no captchas in the information flow, data-rich text over minimalist design. It is the natural evolution of SEO once your visitor is no longer always human.

Do AI crawlers really miss JavaScript-rendered content?

In general they don't execute it. An Adobe analysis covered by Skift in June 2026 showed the practical consequence in travel: crawlers couldn't see fares rendered client-side with JavaScript, and airlines fared worse than hotels and car rental, which serve more content from the server.

Can an AI agent read your website? A 7-point checklist

An AI agent doesn't see your website the way a customer does. It doesn't wait for JavaScript to load, doesn't interpret your design, doesn't wrestle with your captcha: if something gets in its way, it moves on to the next option. And one figure sums up the size of the problem: a hotel's website typically exposes between 5% and 10% of its real operational information, according to an analysis by Mirai (2025) — the rest lives in PDFs, internal documents and systems no external machine can query.

This checklist covers the seven points separating an agent-readable website from an invisible one. You can walk through all of it in an afternoon.

1. Does your content exist without JavaScript?

AI assistant crawlers generally do not execute JavaScript. If your site loads an almost empty page and fills it in with scripts afterwards, that content doesn't exist for them. This isn't theory: an Adobe analysis covered by Skift (June 2026) showed AI crawlers couldn't see client-rendered fares, and that airlines — heavy users of that pattern — fared worse than hotels and car rental companies, which serve more content from the server.

How to check: open your site, hit "View page source" and search for a paragraph you know is published. If it's not there, you have a rendering problem.

2. Are you letting AI bots in?

Your robots.txt file decides which crawlers may read you. Each assistant uses its own, under different names, and many templates and hosting panels ship with restrictive rules by default. Review who is blocked and decide consciously: blocking by default means giving up on appearing in the answers.

3. Are there captchas or walls in the way of the information?

Agents can't solve captchas, as SiteMinder pointed out when describing its AXO (Agent Experience Optimization) concept in 2025. A captcha in front of the availability search, a newsletter interstitial or a badly built cookie wall interrupts the machine's journey like a locked door. Protect submission forms, not public information.

4. Is your key data marked up with structured data?

JSON-LD markup (Schema.org) is the spec sheet a machine reads without interpreting your design: business type, address, phone, services, FAQs, indicative prices. Cloudbeds' GEO checklist (2025-2026) puts it among the technical team's first tasks, alongside server-side rendering and a clean sitemap. On my own website I apply it on every page, together with an llms.txt file: it is hours of work, not months, and it's one of the few things that depends only on you.

5. Does your content answer questions or describe itself?

A machine quotes what it can extract in one sentence. "An oasis of tranquillity by the sea" answers nothing; "we accept dogs up to 20 kg at no extra charge" answers a question someone is asking right now. The practical recommendation circulating in the industry (Cendyn, 2026) is that key facts appear within the first 50-60 words of each block, in question-and-answer format where natural.

6. Is critical information on the web page or in a PDF?

The scanned PDF menu, the conditions in a downloadable document, the opening hours in an image: all of that sits beyond the reach of most crawlers. The rule is simple: whatever you want an AI to be able to quote must be HTML text at a public URL.

7. Have you tested it with an actual AI?

The final test isn't an audit tool: it's asking. Request specific information about your business from ChatGPT, Gemini and Perplexity — opening times, policies, services — and watch what they answer and which sources they cite. If they make things up or only cite third parties, you know which part of your website isn't getting through.

Why this matters now (not "in the future")

This is not about preparing for a hypothetical era. AI referral traffic to US travel websites grew 194% year over year (Adobe Analytics, May 2026, across more than 8 million transactions), with visitors spending 70% more time and bouncing 41% less. It is still a fraction of the total, but it is the fastest-growing, best-behaved traffic there is. Making your website readable by the machines that bring it isn't a bet on the future: it's basic 2026 maintenance.