Can AI
read your website?

Why AI legibility is worth checking

More first impressions now happen inside a chat window. When someone asks ChatGPT, Claude or Google's AI results about a problem you solve, those tools answer from what they can read on the open web, and your website is the main thing they have to go on. If a crawler cannot reach your pages, or reaches them and finds markup with no structure, it falls back on whatever fragments it can find elsewhere, which may be out of date or simply wrong.

Making your site legible does not guarantee a place in anyone's answer, and be wary of anyone who says it does. It gives you the baseline, which ordinary search has always needed too. I have written up how AI search reads your website if you want the longer version of what is going on under the bonnet.

What the checks mean

Three of the checks deserve a word of explanation. Robots access looks at whether your robots.txt file blocks the crawlers that AI tools send out; plenty of sites block them by accident, usually through an old catch-all rule or an over-eager plugin. llms.txt is an emerging convention rather than an official standard, a plain-text file offering AI systems a tidy summary of what your site contains; most sites do not have one yet, so treat a warning there as a nudge, not a problem. Structured data, the JSON-LD in your page's head, is the closest thing the web has to proper labelling: it tells any machine reading the page who you are, what you do and where you are, in a format it cannot misread.

The rest are the basics: one clear h1, a sensible heading order, a title and meta description, a canonical address, Open Graph tags and a sitemap. Passing every check means your site is easy for a machine to read. It does not mean a guaranteed place in an AI answer, and nothing does. If the results here surprise you, the fix is usually straightforward, and it is the kind of thing I build into every site.