ARD check · free

Can AI agents discover your APIs and tools?

Agentic Resource Discovery (ARD) is a proposed standard (spec v0.91) for listing callable resources — APIs, MCP servers, AI agents — in one manifest. This check looks where Lighthouse 13.6 looks and validates the manifest the same way.

Four discovery paths, one validation

1 · Where is the manifest?

Agentmap line in robots.txt, <link rel="ard">, HTTP Link header, then /.well-known/ard.json. The predecessor ai-catalog paths count, but with a warning.

2 · Is it valid?

Required fields per entry (identifier as urn:air:…, displayName, type, exactly one of url or data), media types, 2–5 representativeQueries — the rules of the official conformance test.

3 · Same rating as Lighthouse

Valid = 1, valid with notes = 0.9, invalid = 0. No manifest = “not applicable”, which Lighthouse does not count as an error. Tested against 12 reference pages.

4 · Not a ranking factor

ARD helps agents and registries find callable resources. It does not make AI assistants recommend a website, and nobody can promise that.

To be honest: most websites do not need ARD. If you do not offer an API, MCP server or agent, leave it out — for your content to be found, robots.txt, structured data and llms.txt matter. The ARD result is also part of every GEO audit, separate from the score.