Skip to Content
Improving your Agent Readiness score

Improving your Agent Readiness score

Your Agent Readiness score grades how well your site is set up for AI assistants to find, read, and trust it. It runs a fixed set of checks against a page and gives you a score out of 100, with each failed check explained so you know what to fix. A high score means an assistant can reach your content, parse it cleanly, and treat it as a credible source — the groundwork for being cited. (For why that matters, read What is GEO?; for the mechanics, How AI assistants find and cite your site.)

This guide walks through what the checks look for and how to pass them, grouped by what they are really testing.

Be reachable

If an assistant’s agent cannot get to your page, nothing else counts.

  • Sane crawler rules. Your robots.txt should allow the AI crawlers you want to read your site, not block them by accident. A blanket disallow keeps you out of every answer.
  • Verifiable agents. Well-behaved agents can sign their requests so a site can confirm who they are. Honouring that standard is the strongest reachability signal — though it needs the page to be served from somewhere that can verify a signed request, which a purely static page cannot do on its own.

Be readable

Once an agent has your page, it has to understand it without guessing.

  • Semantic structure. Use real landmarks — header, nav, main, footer — and proper headings. Clear structure with sensible ARIA labels lets an assistant follow your page the way a person would.
  • Structured data. Add JSON-LD describing what the page is. This is the single highest-value thing most sites are missing.
  • The right schema type. Generic structured data is good; the correct type is better. An article should declare itself an Article (or a subtype), a product a Product, and so on, so the assistant knows how to read it.

Be machine-legible

Give assistants a clean, dedicated way to consume your content.

  • An llms.txt file. This is a plain-text guide to your site written for language models — what you are, your key pages, and where to read more. Publishing one is a clear “this site is built for you to read” signal.
  • Feeds and mirrors. Link your llms.txt to your sitemap and to machine-readable versions of your pages, so an agent can discover the whole corpus, not just the page it landed on.

Be fast and light

Agents have budgets too, and a slow or heavy page gets skipped.

  • Quick first byte (TTFB). The page should start responding fast, especially without JavaScript — many fetchers do not run scripts and judge you on the raw response.
  • Reasonable page weight. Keep the total page size modest. A bloated page is expensive to fetch and a poor signal.

Be form-friendly

If your page has forms, label their fields with standard autocomplete values (and the right input types). It helps agents — and people using assistive tech — complete them correctly. A page with no address or payment fields simply passes this check; it does not need to invent forms.

A realistic target

A perfect 100 is not always reachable for every kind of site — the signed-agent check, for example, needs server-side request verification that a static page cannot provide. Do not chase the last point at the cost of the easy ones. Fix structured data, semantic HTML, llms.txt, and speed first; those move the score the most and are exactly what gets you read and cited.

Frequently asked questions

Where do I run the check? On the Agent Readiness tool on the Voris website. Enter a URL and it scores that page and lists what to fix.

Does a higher score guarantee I’ll be cited? No. A high score removes the technical reasons an assistant would skip you — it makes you readable and trustworthy. Being cited still depends on having content that answers the question.

How often should I re-check? After any significant change to your site’s structure, templates, or hosting, and periodically as the rubric evolves. Re-run the check and confirm you have not regressed.

My score dropped after a redesign. Why? Redesigns commonly drop structured data, change robots.txt, or add page weight. Re-run the check and compare the failed checks against what changed.

How AI assistants cite your siteHow to add an llms.txt to your site