What is agentic resource discovery (ARD)? How AI agents will find services, and what it means for your SEO

For years, a simple pattern held: a person searches, Google lists results, the person clicks. That pattern now has competition. The new question is no longer “Which website?” but “Which capability solves my task, and which agent or tool provides it?”. On June 17, 2026, Google introduced Agentic Resource Discovery (ARD), a standard that formalizes exactly this shift. And it is more than a Google project.

What is agentic resource discovery (ARD)?

Agentic Resource Discovery is an open standard “to publish, discover, and verify AI capabilities across the web”. The specification was written by Junjie Bu and Srinivas Krishnan. It is published under the Apache 2.0 license and builds on the “AI Catalog data model” of the Linux Foundation AI Catalog Working Group.

ARD answers three questions: Where does the right capability live? Which one should I use? And how do I verify securely that I am talking to the right source? Besides Google, the early supporters include Snowflake, Hugging Face, Microsoft and Cisco (Outshift). Google integrates ARD into the Gemini Enterprise Agent Platform through an “Agent Registry”; according to Google, native support will follow “in the coming months”.

The specification has two core building blocks:

• Catalogs: an organization publishes a catalog of its capabilities on its own domain. Domain ownership forms the cryptographic foundation of trust.
• Registries: they act as “search engines for the agentic web”. They crawl the catalogs, index them and return matching results to agents, together with verification metadata.

How ARD works: how agents discover resources compared with search engines

The process has four phases:

1. Publish a catalog: you place a manifest at /.well-known/ai-catalog.json on your own domain (or alternatively through robots.txt directives, HTML link tags or DNS service bindings). Each entry gets a domain-anchored URN following the pattern urn:ai:<publisher>:<namespace>:<agent-name>.
2. Discovery & resolution: a registry crawls the manifest. An agent then sends a natural-language query via POST /search; the registry returns results sorted by relevance, with a score from 0 to 100.
3. Cryptographic verification: optionally, a trustManifest comes into play, for example a SPIFFE ID, a DID or an HTTPS FQDN, plus attestations such as SOC2 or HIPAA. The key point: the trust domain must match the domain in the discovery identifier.
4. Direct runtime connection: only after a successful check does the agent connect directly to the capability.

What ARD means for SEO and online marketing

We see ARD as the start of “discovery optimization”, an extension of SEO and GEO. Three consequences are emerging:

• Your domain becomes the anchor of trust. Because the foundation of trust depends on domain ownership, in this model it partly replaces classic backlink authority.
• Description quality becomes a ranking factor. The specification explicitly provides for representativeQueries (2–5 of them) and a description, from which registries build “semantic vector embeddings for search ranking”. If you write precisely here, you are easier to find.
• Disintermediation and governance. If an agent calls a capability directly, the classic click may disappear entirely. At the same time, the trustManifest opens up a new advisory field around governance and compliance.

A realistic assessment: limits and open questions

To be honest: at the time of writing, ARD is just one day old. According to Google, native Gemini support will only come “in the coming months”, and whether the standard will catch on is an open question. Still, the early list of supporters is unusually strong for such a young standard.

There is no need for hasty action, but the direction is clear: away from “a person looks for a page” toward “an agent looks for a capability”. In practice, that means getting ready instead of rushing: check how findable you are now, rather than catching up later.

Conclusion

ARD belongs on your radar, not as an emergency drill but as an early strategic signal. If you treat your domain as an anchor of trust and write your description texts carefully, you are laying the groundwork today for tomorrow's agentic search. Our discovery optimization service shows how to turn this into a concrete plan: from an ARD readiness audit and the ai-catalog.json to monitoring. Closely related is our view on the agent readiness of your website: how a site becomes findable, understandable and usable for AI agents in the first place.

At its core, ARD moves the logic of findability further away from the blue link and toward a verifiable, machine-readable capability. It is the same basic shift that drives GEO optimization and that determines how AI search systems select sources. If you take it seriously early on, you secure your visibility in the next generation of search.

Sources: Junjie Bu & Srinivas Krishnan, “Announcing the Agentic Resource Discovery specification”, Google Developers Blog, June 17, 2026, developers.googleblog.com; technical details from the ARD specification, agenticresourcediscovery.org/spec.