AI for Business
What Is Agentic Resource Discovery (ARD)? A Guide for Business Owners
Agentic Resource Discovery (ARD) is a new open standard that lets AI agents find tools, services, and other agents the way regular search helps people find websites. It was announced in mid-June 2026 — not by Google alone, as some early coverage suggested, but by a coalition of around eleven companies including Google, Microsoft, and Hugging Face, under a Linux Foundation working group, and released as a free, open specification. The short version for a business owner: the web is quietly splitting into two layers, one built for humans and one built for machines, and ARD is the directory for the machine layer. Here's what it is, how it works, and whether you need to do anything about it yet (for most businesses, the honest answer is "not today, but soon").
What Agentic Resource Discovery actually does
Regular search helps a person find a web page. ARD does something different: it helps an AI agent find a capability it can use — a tool to call, an API to query, or another agent it can hand part of a task to.
A quick example. Say an agent needs to check the keyword density on a page and doesn't already have that ability built in. Instead of a developer wiring that tool in ahead of time, the agent can describe what it needs in plain language, get back a matching tool, confirm who published it, and start using it on the spot. No pre-arranged integration. That "find it at the moment you need it" step is the whole point of ARD.
It helps to think of ARD as a search engine for agents. The same way Google crawls and ranks web pages so people can find them, ARD-style registries crawl and rank machine capabilities so agents can find them.
How ARD works: catalogs and registries
The standard rests on two simple pieces.
A catalog is a file you host on your own domain at a predictable address — yourdomain.com/.well-known/ai-catalog.json. It's a machine-readable menu of everything you offer to AI agents: each entry names a capability, describes what it does, says how an agent should call it, and lists a few example questions it can answer.
A registry is the search-engine part. Registries crawl these catalog files across the web, index every capability they find, and answer agent requests. When an agent needs something, it sends a plain-language request to a registry, which matches it against the example questions in everyone's catalogs, ranks the results, and returns the best matches along with proof of who owns each one.
Put together, the flow is four steps:
- Publish. You host your ai-catalog.json file on your domain.
- Discover. An agent either queries a registry in plain language, or — if it already knows you — fetches your catalog directly.
- Verify. The agent confirms who published the capability. Because the catalog lives on your own verified domain, an agent can trust that a tool listed under your domain really is yours.
- Connect. The agent calls your tool through its normal protocol and gets to work.
Trust is anchored to domain ownership, the same principle that makes a verified domain meaningful for email today. The spec layers on additional checks too — things like compliance attestations and cryptographic identity — but for most owners, "the capability is provably tied to your domain" is the idea that matters.
ARD vs SEO: how they differ, and why you need both
SEO optimizes your pages so people find you. ARD exposes your capabilities so AI agents find them. Different users, different mechanics.
Traditional SEO is about ranking web pages for human queries — you tune titles, content, and structured data so a person clicks through. ARD has no ranking algorithm in the SEO sense. An agent picks a capability based on how well your example questions match its request, and on whether it can verify you. You're optimizing for machine readability: clear endpoints, clear descriptions, accurate metadata.
Here's how the two line up, point by point:
- Who it serves — SEO serves human searchers; ARD serves AI agents.
- What it exposes — SEO exposes web pages and content; ARD exposes callable tools, APIs, and agents.
- How you get found — with SEO, by ranking in search results; with ARD, by matching an agent's plain-language request.
- The file involved — SEO relies on sitemap.xml; ARD relies on ai-catalog.json.
- What you optimize — SEO: titles, content, and links; ARD: descriptions, example questions, and endpoints.
One row there does a lot of work. The example questions you write for each capability act much like a title tag does in SEO — they're the text an agent matches against, so a vague description gets your tool skipped the same way a vague title buries a page.
These two approaches don't compete. People will keep finding you through search for a long time yet, and machine readability sits alongside that, adding a second way to be found. It's an additional channel, not a replacement.
Why this is happening now
The shift underneath ARD is simple: AI agents are moving to the center of how people get things done. Increasingly, instead of visiting several sites, reading each one, and filling out forms yourself, you'll describe a goal to an agent and let it do the visiting, reading, and contacting on your behalf.
A growing share of shoppers already lean on AI to research and compare before they buy, and that number keeps climbing. As it does, the agent effectively becomes the buyer — and when the buyer is a machine, your business has to be readable by a machine, because a person is no longer the one clicking through your site. The companies building the agent ecosystem saw this coming and are laying the infrastructure for it. The businesses whose capabilities show up when an agent goes looking are the ones that get used.
Should your business set up ARD right now?
For most small and mid-size businesses, the honest answer is: not yet — and here's the reasoning, because it matters.
ARD is only useful today if you actually host tools or agents that other agents can use. Most businesses don't, so there's nothing to put in a catalog. Publishing an empty menu accomplishes nothing.
It's also genuinely early. The specification is a draft, and adoption is currently near zero — as of mid-June 2026, a survey of major sites, including the companies that authored the standard, found that essentially none had published a discoverable catalog yet. This is a forecast to watch and a cheap bet to consider, not a proven channel.
That said, two things are changing fast. AI has made it realistic for an ordinary business to stand up a working tool or agent, which wasn't true even a year ago. And there's a real first-mover advantage: the file itself is small and cheap to publish, so if a competitor exposes a capability an agent can find and you don't, the agent defaults to them.
So the sensible move for most owners is to understand the shift now and get the foundation in place — which means making your website readable and usable by agents first. The catalog and tools come after you have something worth exposing.
What to do about ARD right now
Start one layer below ARD. Make your website something an AI agent can actually read, understand, and act on: clean structure, accurate descriptions of what you do and who you serve, honest and current information, and machine-readable data wherever it helps. That work pays off no matter how quickly the tool-and-agent layer arrives, and it's the foundation everything else sits on.
Then keep an eye on the milestone that will tell you ARD has arrived for real: the first time a mainstream AI assistant queries a registry by default. Until then, you're building readiness, not chasing a live channel.
Common questions about ARD
Is ARD the same as MCP? No. MCP gives an agent a standard way to call a tool it already knows about. ARD is the layer in front of that — it helps an agent find the right tool in the first place. ARD is the search step; MCP is the connection step.
Isn't this just a Google thing? No. Some early coverage framed it as "Google's search engine for AI agents," but ARD is an open standard co-authored by Google, Microsoft, and Hugging Face and backed by a working group of around eleven companies under the Linux Foundation. It's released under an open license, so anyone can run a registry or build an agent that uses it.
Does ARD replace SEO? No. SEO gets your pages in front of people; ARD exposes your capabilities to agents. They handle two different jobs and work best together.
Do I need to set up ARD today? Only if you already host tools or agents that other agents can use. Most small businesses don't yet, so the first step is making your website readable and usable by agents. The catalog comes once you have a working capability worth listing.
At Workhorse Solutions, we help small and mid-size businesses get ready for this shift — starting with the foundation, your website's readability to both people and agents, and moving up from there as the tool layer matures. Book a free consultation and we'll show you straight where your site stands and what's worth doing next.