Agent Readability Score

What does an agent actually get from your page?

Paste any public https URL. We fetch it twice — once with an agent’s Accept header, once with a browser’s — and score the result against the published spec. Free, no account, and every check shows the evidence it was scored on.

How it is scored

One hundred points, the same six dimensions for every page

No per-vertical weights: an 88 means the same thing on a product page, a clinic’s hours page and a government form. What varies is which facts that kind of page is expected to carry, not what they are worth.

  • Retrievability

    20

    2xx on the agent path, robots policy, render independence

  • Machine representation

    20

    content negotiation, declared alternates, Vary, parity

  • Fact coverage

    20

    core facts for this page kind, and what they cost in bytes

  • Fact position

    10

    how deep into the response the first core fact appears

  • Structured data

    15

    JSON-LD/microdata present, typed, complete, and agreeing with the text

  • Contract & discovery

    15

    canonical, cache validators, llms.txt, sitemap, declared endpoints

64 of those 100 points are measured and 36 are heuristic. Every heuristic check is labelled as one wherever it appears, with the inputs it was inferred from — a number you cannot audit is a number you should not act on.

Before you run one

What this measures, and who wrote it

Both of these print on every result too. They are not a footnote.

ARS measures format and retrievability. It does not measure whether the facts are true, or whether any assistant cites this page.

ARS is authored by Rebilder, which sells content-negotiation middleware. Two ARS checks (D2.1 content negotiation, D5 structured data) can be satisfied by Rebilder's free gateway, and reaching an A requires both. The full ruleset, thresholds, and conformance corpus are published so this bias is auditable rather than hidden.

Scanning is not publication. Anyone can scan any public URL and see the result, and nothing about a domain we do not have verified owner opt-in for is ever listed, ranked, badged, or indexed. Result pages for unclaimed domains carry noindex and no preview card. Removal from anything we do publish is free, self-serve and permanent.

We identify honestly as rebilder-ars/0.1 on every request, obey robots.txt, and never wear another crawler’s name. Crawler policy, request budget and opt-out.