Methodology

Readiness levels and scoring

A level is a cumulative statement about what an agent can already do with your site. Rungs are evaluated in order and a site sits at the highest rung it fully satisfies.

The five levels

Levels are cumulative: a site sits at the highest rung where every requirement is fully satisfied. Reaching Level 3 requires Level 2 first.

0

Not Ready

No crawl policy or discovery basics are in place yet.

1

Basic Web Presence

Pass two of three: robots.txt, a sitemap, or Link headers.

  • robots.txt
  • Sitemap
  • Link headers
2

Bot-Aware

Level 1 plus explicit AI bot rules and Content Signals in robots.txt.

  • AI bot rules
  • Content Signals
3

Agent-Readable

Level 2 plus markdown content negotiation, so agents read clean content.

  • Markdown negotiation
4

Agent-Integrated

Level 3 plus at least one discoverable integration: an MCP Server Card, A2A Agent Card, Agent Skills index or API Catalog.

  • MCP Server Card
  • API Catalog
  • Agent Skills
5

Agent-Native

Level 4 plus two of three: verifiable bot identity (Web Bot Auth), integrations in place, and auth metadata (OAuth or auth.md).

  • Web Bot Auth
  • OAuth discovery
  • Auth.md

Partial scans

You can choose which checks to run. A check that was not run never blocks a level, because it is skipped rather than failed, so a focused seven-check scan of a well-built content site reports a high level. The result card always labels a partial scan with the exact count, so a high number is never mistaken for a full pass.

Scoring

The 0-100 score is the share of applicable checks that were satisfied. Checks excluded by configuration, or not applicable to this kind of site, are removed from both the numerator and the denominator; a warning counts as half credit.

How the level criteria were derived

The ladder and the pass criteria were reverse-engineered from observed behaviour of the reference scanner across a sample of sites, then pinned down with regression tests. Where a criterion was ambiguous, the interpretation that reproduces the reference result was chosen and is documented in the tests.

Excluded checks

Web Bot Auth is informational: publishing a key directory claims that you sign your outbound requests, so a site that does not sign should not publish one. The five commerce checks are scored only on sites with e-commerce signals.

Every check, its category and whether it is enabled by default
CheckCategoryDefaultScored on
robots.txt Discoverability yes every site
Sitemap Discoverability yes every site
Link headers Discoverability yes every site
DNS for AI Discovery Discoverability yes every site
Markdown negotiation Content yes every site
AI bot rules Bot Access Control yes every site
Content Signals Bot Access Control yes every site
Web Bot Auth Bot Access Control yes every site (informational)
API Catalog API, Auth, MCP & Skill Discovery yes every site
OAuth discovery API, Auth, MCP & Skill Discovery yes every site
OAuth Protected Resource API, Auth, MCP & Skill Discovery yes every site
Auth.md API, Auth, MCP & Skill Discovery yes every site
MCP Server Card API, Auth, MCP & Skill Discovery yes every site
A2A Agent Card API, Auth, MCP & Skill Discovery opt-in every site
Agent Skills API, Auth, MCP & Skill Discovery yes every site
WebMCP API, Auth, MCP & Skill Discovery yes every site
ARD manifest API, Auth, MCP & Skill Discovery yes every site
x402 Commerce yes commerce sites only
MPP Commerce yes commerce sites only
UCP Commerce yes commerce sites only
ACP Commerce yes commerce sites only
AP2 Commerce opt-in commerce sites only

See it on your own domain

The scanner shows its evidence for every check, so you can audit the result.

Run a scan