On June 17, Google announced Agentic Resource Discovery (ARD), an open specification it describes as "An open specification for finding and verifying tools, skills, and agents across the web." Most coverage treated it as a developer story. It is not. It is a story about where trust lives on the web that is forming around AI agents, and the answer Google gives is: your domain.
What Google announced
ARD standardizes how AI capabilities get published, found, and verified across the open web. The spec has two building blocks.
Catalogs. An organization publishes a file called ai-catalog.json at a well-known path on its own website. The catalog lists what the organization offers to agents: MCP servers, A2A agents, API tools, even other catalogs.
Registries. Registries crawl those catalogs and index them, the way search engines crawl pages. Google's own words:
Registries act as search engines for the agentic web. They crawl published catalogs, index their contents, and make them searchable.
When an agent needs a capability, it asks a registry in plain language, gets back matching results, verifies who published each one, and connects directly.
The part worth reading twice is how ARD anchors trust:
Because these catalogs are hosted directly under the organization's own domain, ownership of that domain serves as the cryptographic foundation for identity and trust.
In plain English
Google is answering three questions agents keep running into, which the announcement states outright: where does the right capability live, which one should the agent use, and how does it verify the capability is safe to connect to.
Today there is no standard answer. Every platform keeps its own private list of tools and agents, and none of those lists talk to each other. ARD replaces the private lists with a public pattern:
- You describe what your business offers to agents in one machine-readable file.
- You host that file on your own domain, which proves the offer is really yours.
- Registries index it so any agent, on any platform, can find it.
If that pattern sounds familiar, it should. It is how the searchable web was built: publish on your domain, get crawled, get found. ARD applies the same architecture to agents, and Google draws the comparison itself, saying that "Just as the open web democratized information, ARD democratizes AI resource discovery."
Why this matters for your business
Three practical consequences fall out of this spec.
Your domain is the asset. Not your listing on someone's platform, not your profile in someone's registry. ARD puts identity and trust at the domain level. The businesses that treat their domain as an agent-ready property will be discoverable in this new index. The ones that don't will not exist in it.
Discovery is moving below the page. A registry doesn't read your homepage copy. It reads your catalog: a structured declaration of what you offer and how to call it. Being legible to humans and being legible to agents are now two separate jobs, and most websites only do the first one.
Verification is not optional. ARD lets publishers attach verifiable trust metadata so an agent can confirm who it is talking to before it connects. Unverifiable capabilities are exactly what registries and client agents are being built to filter out.
Where Trinzik fits
This spec describes the architecture we already build. Our Apex Domain Agents work turns a client's domain into exactly the kind of property ARD assumes: a website that is the source of truth, an AI agent grounded in that website's content, machine-readable discovery files, and an agent card signed with a cryptographic key so other systems can verify it before they trust it. Our own site runs this way today; the agent card at trinzik.ai/.well-known/agent-card.json is signed, and anyone can check the signature against our published keys.
We built it that way because discovery without verification is noise. Google standardizing the same posture, with domain ownership as the root of trust, tells you which direction the agentic web is settling.
If you want to know what your domain looks like to an agent right now, that is a short conversation. Book a walkthrough and we will show you.
Questions this raises
What is the Agentic Resource Discovery (ARD) specification?
Agentic Resource Discovery (ARD) is an open specification, announced by Google in June 2026, for publishing, discovering, and verifying AI capabilities across the web. An organization lists what it offers to agents in an ai-catalog.json file hosted on its own domain, and registries crawl those catalogs and make them searchable, the way search engines index web pages.
Does ARD replace the A2A agent card?
No. ARD is a discovery layer that sits above individual agents. A catalog's entries can include A2A agents, MCP servers, and API tools, so the agent card still describes and identifies a single agent; ARD is how other systems find that agent across organizational boundaries and verify who published it before connecting.
Why does hosting the AI catalog on your own domain matter?
Domain ownership is ARD's trust anchor. The specification treats ownership of the domain as the cryptographic foundation for identity and trust, which means a capability published under your domain can be verified as genuinely yours before an agent connects to it. A business that controls its domain controls its agent identity.
Sources
- Announcing the Agentic Resource Discovery specification · Google for Developers Blog, June 17, 2026
- Agentic Resource Discovery specification (ard-spec) on GitHub