Google, Microsoft Back Draft AI Agent Discovery Spec

Eleven companies, including Google, Microsoft, GitHub, and Hugging Face, have published Agentic Resource Discovery (ARD). The open specification sets out how AI agents find and verify tools, skills, and other agents across the web.

The contributors released the draft spec on June 17, along with reference implementations from several of them. ARD is licensed under Apache 2.0 and builds on the AI Catalog data model maintained by a working group under the Linux Foundation.

What Does ARD Solve?

The spec sets out to solve a coordination problem. Today, an agent has to be wired to each tool, MCP server, or API it uses ahead of time. As more companies publish their own capabilities, that pre wiring stops scaling. ARD moves discovery. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

The risk is usually hidden in the execution layer. A page can look fine to a human and still fail for an automated visitor if the form, call to action, rendering path, or confirmation step is not accessible enough for the agent to complete the task.

How ARD Works

ARD relies on two pieces, which the spec calls catalogs and registries. An organization publishes a catalog, an ai catalog.json file hosted at a well known path on its own domain, that lists the tools, MCP servers, agents, or APIs it makes. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

Same Day Implementations

Several contributors shipped working tools built on the spec the same day. GitHub introduced agent finder, which lets Copilot discover matching MCP servers, skills, tools, and agents from a chosen registry, with users controlling what gets. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

Where Google Fits

Google's role centers on Agent Registry, part of its Gemini Enterprise Agent Platform. The company said Agent Registry will host and search agentic resources and handle enterprise governance. Native ARD support in the platform is planned. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

Why This Matters

The split depends on what you publish. ARD is for publishers of callable capabilities, the APIs, MCP servers, and agents that software connects to. A company that publishes tools has a clear method for being found and trusted by agents. A. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

Looking Ahead

The spec is a v0.9 draft, and the contributors are inviting changes through the project's GitHub repository. Its reach depends on registries that can crawl and index catalogs at scale, and that ecosystem is still in its early stages. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

What Does ARD Solve? in practice

Introduction Eleven companies, including Google, Microsoft, GitHub, and Hugging Face, have published Agentic Resource Discovery (ARD). The open specification sets out how AI agents find and verify tools, skills, and other agents across the. The strategic issue is whether automated visitors can understand, trust, and complete the same journey a human visitor can. Agent readiness is partly technical, but it is also about clear tasks, accessible flows, and reliable evidence.

What the visibility signal actually changes

What the visibility signal actually changes: google, Microsoft Back Draft AI Agent Discovery Spec: the Practical Angle should be treated as a visibility signal, not a standalone headline. Introduction Eleven companies, including Google, Microsoft, GitHub, and Hugging Face, have published Agentic Resource Discovery (ARD). The open specification sets out how AI agents find and verify tools, skills, and other agents across the web. The. This connects with 4 Layer AI Ops Playbook when the same signal needs a clearer operating decision. A useful companion note is Google Says Markdown, because it looks at a nearby part of the same system.

What the visibility signal actually changes: the practical question is whether the page, brand evidence, and surrounding content make the answer easier to trust. If that support is weak, search systems can still understand the topic but fail to connect it confidently to the brand.

What the visibility signal actually changes: that is why the response should begin with an audit of the evidence already on the site before creating a new asset. The fastest improvement is often a clearer page, a better internal link, or a stronger explanation of why the brand belongs in the answer. The same pattern also shows up in Written for Readers Who Don’t Read, where the practical question is how the signal becomes visible.

Where the evidence needs to be tested

Where the evidence needs to be tested: a single study or ranking observation should not become a strategy by itself. It should become a diagnostic prompt: which source is being trusted, which query pattern is affected, and which part of the site would make that trust easier to earn?

Where the evidence needs to be tested: that keeps the response grounded. The goal is to improve the evidence chain around the topic rather than publish another summary that repeats what every other page already says.

Where the evidence needs to be tested: the important distinction is between a useful signal and a fashionable talking point. A useful signal changes the brief, the page structure, the linking plan, or the measurement view.

How to avoid overreacting to one data point

How to avoid overreacting to one data point: for content teams, the strongest move is to map the claim to existing assets before creating anything new. The right page may already exist, but it may need clearer headings, stronger internal links, fresher proof, or a better explanation of why the brand belongs in the answer.

How to avoid overreacting to one data point: this is also where title rewriting matters. A title should not copy the source headline; it should frame the practical implication so readers immediately know why the topic deserves attention.

How to avoid overreacting to one data point: the same standard should apply to every section. Each heading needs to earn its place by moving the reader through the evidence, not by repeating the outline in a more polished voice.

What this means for content and authority

What this means for content and authority: authority is becoming more contextual. It is not enough to be generally known in a category if the specific answer depends on a different source, a different index, or a different retrieval pattern.

What this means for content and authority: that means the content system should show consistent entities, related pages, credible references, and useful depth around the exact questions people and AI tools are asking.

What this means for content and authority: when the context is weak, AI systems can still mention the brand but describe it in the wrong frame. The fix is not more volume; it is cleaner evidence around the specific association.