Structured data and llms.txt: being read by AI, not just Google
AI agents and assistants now visit websites on behalf of people. What your site needs so they understand it, cite it, and send you customers.
More and more "visitors" to your site aren't people — they're AI agents reading on behalf of a user: "find me a studio to build a store", "call to book an appointment". If your site isn't machine-readable, those agents skip you.
Three levels of machine readability
1. Clean HTML (non-negotiable)
If a page renders only with JavaScript and the text isn't in the source HTML, many bots see a blank page. Server-rendered content, real headings, links as <a href>, contact details in plain text (not only in a script).
2. Structured data (Schema.org)
JSON-LD blocks that tell machines what this is:
Organization/LocalBusiness— who you are, where, phone, hoursProduct/Offer— price, availability, ratingsFAQPage— questions and answers (Google often shows them directly)Article/BlogPosting— author, date, topicBreadcrumbList— where the page sits in the structure
Check with Google's Rich Results Test.
3. llms.txt
A short file at https://your-site.com/llms.txt, written in the first person: who you are, what you offer, links to the key pages, contact. Still a new convention, but cheap to do and already read by some assistants. The robots.txt analog, for AI.
Why it's worth doing now
- AI answers cite sources — being machine-readable is the new "being found".
- Agents that do tasks on behalf of people are growing fast. Those who are ready get referred.
- All of this also helps classic Google SEO — it's not either/or.
What NOT to do
- Structured data that lies (marked-up reviews you don't have) — Google penalizes it.
llms.txtfull of marketing fluff. Facts, links, contact.
Want a site both Google and AI agents understand? All of this is standard with us — write to us and you get it built in, not as an add-on.