Can AI read your website? Here's how to find out.

The quiet change in how buyers find you

For twenty years the deal was simple: people typed words into a search box, a list of links came back, and your job was to be on the list. That deal has changed. A growing share of your buyers now ask a question in plain language and get a plain-language answer. No list. No click. Just an answer, with your business either in it or not.

Ask ChatGPT, Claude, or Perplexity the question your best client asked before they found you. Something like "who helps credentialing organizations fix their operations" or "best fractional COO for a founder-led services firm." Read what comes back. That answer was assembled from what those engines could read about you and your competitors. If your site is hard to read, the engine works with less, and it either skips you or guesses.

The engines are already answering questions about your business. The only open question is what you've given them to work with.

What machines actually read

A machine reading your website doesn't see the design. It sees the underlying document: the HTML. It reads that document top to bottom, looking for structure that tells it what things mean, not just how they look. A human sees a beautiful page and infers that the big bold line is the point. A machine only knows the big bold line is the point if the document says so.

This is why two sites that look equally good to a person can be wildly different to a machine. One is a well-marked building: every room labeled, a directory in the lobby. The other is the same building with the labels torn off. Same rooms, same contents, but a visitor who can't see has no way to move through it.

The five signals that decide it

• 1. The content is in the HTML, not painted on after. Some sites ship a nearly empty document and use JavaScript to fill in the words once the page loads in a browser. People never notice. Machines often do, because many crawlers read the document as delivered and move on. If your real content arrives late, to a machine your site is a lobby with the lights off. The check: view your page's source (right-click, "View Page Source") and search for a sentence from the middle of the page. If it's not there, machines may not be seeing it either.
• 2. The structure is semantic. HTML has elements that carry meaning: one <h1> that names the page, headings that step down in order, <main> around the primary content, <nav> and <footer> where they belong. Pages built as a thousand anonymous boxes make a machine work for meaning it shouldn't have to work for. The check: does each page have exactly one H1, and does it say what the page is about?
• 3. Structured data describes the business. Schema markup (JSON-LD) is a small block of code that states plainly: this is a business, this is what it sells, this is what it costs, these are the questions people ask and the answers. It's the difference between a machine inferring what you do and a machine being told. The check: paste your URL into Google's Rich Results Test and see what it finds.
• 4. The crawl doors are open. Your robots.txt file tells crawlers what they may read. Some sites block AI crawlers without knowing it, often from a template someone copied years ago. An llms.txt file goes further: a clean, current, plain-text summary of the business, written for machines to read first. The check: visit yoursite.com/robots.txt and yoursite.com/llms.txt and see what's there.
• 5. The words carry the real story. None of the structure matters if the content is stale. If your site still describes the 2022 version of your business, machines will confidently repeat the 2022 version to your 2026 buyers. Machines don't fact-check you against reality. They repeat what you published.

The ten-minute check

1. Ask the engines your buyer's question. Open two or three AI engines and ask what your prospects ask. Note what's wrong, what's missing, and who gets named instead of you.
2. View source and search for your own sentence. If the words on the page aren't in the document, you have a rendering problem.
3. Check robots.txt and llms.txt. One minute each. You're looking for accidental blocks and for whether the machine front door exists at all.
4. Run one page through a structured-data checker. Google's Rich Results Test is free. No results found means machines are inferring everything about you.
5. Read your homepage as a stranger. Would someone who knows nothing about you learn who you serve, what you sell, and what it costs? If a person can't, a machine can't either.

The mistakes that undo good sites

Some of the least readable sites belong to businesses that spent real money on them. The same handful of mistakes shows up again and again:

• Blocking AI crawlers by accident. A security plugin, a copied robots.txt, or a well-meaning developer blanket-blocks bots — and with them GPTBot, ClaudeBot, and PerplexityBot. The business is invisible to AI engines by its own instruction and nobody remembers writing it.
• Trapping the good content in PDFs. The methodology, the standard, the pricing sheet — the substance of the business — lives in downloadable PDFs while the pages hold marketing gloss. Machines index pages far better than they index attachments. If your best thinking is a download, your best thinking is invisible.
• Publishing words as pictures. Pricing tables exported as images, quotes as graphics, headlines baked into hero images. A person reads them fine. To a machine, an image without alt text is a blank space exactly where your most important claims were.
• Letting the answers live only in a chatbot or a form. If the only way to learn what you charge is to "book a call," machines learn nothing, and increasingly your buyers ask machines first. Whatever you make people ask for, machines will never cite.

None of these are exotic. All of them are invisible from the founder's chair, because the site looks perfect in a browser. That's the trap: machine readability fails silently.

What this looks like when it works

A real example from my own client work: a certification body whose entire standard — the substance of the business — lived in a 300-page PDF. Rigorous, credible, and completely unreadable to machines. Search engines saw a brochure site. AI engines asked about the certification either guessed or said nothing.

We moved the standard out of the PDF and into structured, machine-readable pages: every requirement addressable, every section linkable, schema describing what the certification is and who it's for. Nothing about the standard changed. Everything about its findability did. The same substance, republished in a form machines could read, started showing up in the answers its buyers were already asking.

That's the general shape of the win: you rarely need new content to become AI-readable. You need the content you already paid to create, freed from the formats that hide it.

What a good result looks like

A machine-readable site isn't a technical trophy. It shows up in the business: AI engines describe you accurately and cite you. Search engines pull your actual answers into results. Prospects arrive already knowing what you do and what it costs, because the machine that briefed them was working from a clean document.

And there's a compounding effect. Every new page you publish on a readable site inherits the structure. On an unreadable site, every new page inherits the problems. That's why this is worth fixing before you invest in content, not after.