Google Lighthouse 13.5 Adds AI Agent Resource Discovery Audit

Google's Lighthouse 13.5 has introduced an audit for Agentic Resource Discovery (ARD), a proposed specification designed to help AI agents find the tools and services an organization offers. The update was reported by Search Engine Journal on September 21, 2026.

According to Search Engine Journal, the audit compares a site's ARD catalog against the schema defined by the ARD project. If a site does not provide a catalog pointer, Lighthouse defaults to checking /.well-known/ai-catalog.json. The Lighthouse team expects the update to roll out in Chrome 156 DevTools, and in PageSpeed Insights within two weeks of release.

The audit sits within Lighthouse's experimental Agentic Browsing category and is separate from SEO audits. Search Engine Journal notes it "is not tied to Google Search anywhere in the release notes." Reports will display an ARD result or "Not Applicable" if no catalog is found.

What the Audit Checks

In the 13.5 source code, Lighthouse checks the robots.txt file for an Agentmap line — a directive from the ARD specification pointing to a catalog. It also looks for a link tag with the ai-catalog relation and checks the HTTP Link header for the same relation. If none of the three are found, it requests /.well-known/ai-catalog.json. Schema errors or a pointer to an unloadable catalog cause the audit to fail.

The new audit is grouped with the existing llms.txt audit under a heading called "Agent Discoverability." Google's documentation states that the Agentic Browsing category does not produce a 0–100 score, instead showing a pass ratio, because "the standards for the agentic web are still emerging."

Specification File Renamed

The ARD specification has undergone updates since the audit was built. ARD v0.91 now specifies /.well-known/ard.json as the manifest location, with ai-catalog.json designated as the earlier path. Software reading these files is required to fetch ard.json and can also check the older path. The spec cautions that a file only available at the old path "may not be found."

Search Engine Journal reports it did not find any references to ard.json or the ard link relation in the 13.5 source code or in the main branch of the project as of September 21. The spec also mentions discovery through in-page JSON-LD and DNS records, but the 13.5 code does not check either. Lighthouse built its validator based on the ARD project's own conformance tests.

ARD's Relationship to Other Agent Standards

ARD is distinct from related standards. According to the source, ARD focuses on discovering resources — including MCP tools, A2A agents, skills, and other callable services — so AI systems can find them through search. An llms.txt file, by contrast, summarizes a site's content for agents. WebMCP allows a page to offer structured actions that an agent can call once it is on the site.

Authorship and Status

The ARD specification is still a proposal. It credits three authors: Junjie Bu from Google, R.V. Guha from Microsoft, and Shaun Smith from Hugging Face. The release also adds a weekly automated check that flags when the ARD project's schema or conformance test changes upstream.