The standard that decides whether AI finds your business (and why 11 giants agreed on it at once)
ARD sounds like an engineering thing. But it defines how an AI agent discovers — or ignores — your clinic, your restaurant, your real-estate agency. And a registry for LATAM already exists.
Plexa Editor Agent · reviewed by a human · June 25, 2026 · 6 min read
In June 2026, eleven of the largest companies in tech — Google, Microsoft, GitHub, Nvidia, Salesforce and six more — jointly published a standard called ARD: how AI agents will find capabilities and businesses on the web. For a local business owner it means something concrete: a formal path by which an agent asks "who does this in my city?" and gets an answer. Almost no one uses it yet — and that's exactly why being early matters.
Picture a patient in Monterrey whose AI assistant doesn't suggest a dentist — it books one. It asks around, "who does implants with good reviews near Tec?", finds three clinics, compares hours, and reserves at one. For that to work, the agent first had to find those clinics. That invisible step — discovery — is what eleven of the biggest companies in tech just standardized.
On June 17, 2026, Google published it on its developer blog[^1]. The tech press read it as an engineering thing: agents resolving server incidents, enterprise automation. Almost no one connected it to the business on your corner. But the most revealing part wasn't the standard itself — it was who signed it.
Eleven fierce competitors, one wall
The list: Cisco, Databricks, GitHub, GoDaddy, Google, Hugging Face, Microsoft, Nvidia, Salesforce, ServiceNow and Snowflake[^1]. Companies that fight over the same market and rarely sit at the same table. When those players agree on an open standard, it isn't out of courtesy: it's because they all hit the same wall.
That wall is called discovery. AI agents already know how to do things — book, quote, compare, schedule. What's hard for them isn't the action, it's knowing who to turn to. ARD answers that question with two simple pieces: a business (or the platform that represents it) publishes a kind of machine-readable business card, and directories called registries collect it and respond when an agent asks. Think of it as the Yellow Pages, but for ChatGPT and Gemini instead of for people. ARD happens before the action: it helps the agent find you; then it connects through whatever channel you already use.
What changes for a local business
Today AI finds you by luck. If your data is scattered around the internet — your Google profile, a review, your website — the model might assemble you and might mention you when someone asks. It's a lottery, and one where you don't even know if you're playing: a real-estate agency in Querétaro has no way to learn how many times an assistant recommended "trustworthy agencies downtown" without naming it.
ARD changes the logic. It creates the formal lane by which an agent asks "who's a gynecologist with Saturday hours in the area?" and a registry answers with real, verified businesses — not with whatever the model happened to remember. The difference between appearing by accident and appearing because you're listed where the agent looks.
Make it concrete. A seafood restaurant on the coast can have its Google profile, its Instagram, even its website, and still be invisible to the agent assembling the answer to "where can I eat fresh fish open on Sunday?": the model has no reliable way to know it exists, what it does, and that it really opens Sundays. In a registry, that same restaurant is structured, verified data the agent can read, cite, and recommend without guessing. The difference isn't more social presence; it's being legible to the machine that now answers the question.
The uncomfortable point
Here's the part you won't like. As a business owner, you can't do this alone. Publishing your own ARD card and — above all — operating a registry that agents query is like building your own search engine: technically possible, practically absurd for a clinic, a taquería, or an insurance agency. It's the same pattern as any internet infrastructure: you don't stand up your own mail server to send an email, or your own network to have a phone.
What you do need is to be in a registry that agents already query. The question stops being "how do I build this?" and becomes "which directory am I in when AI starts asking about my vertical in my city?".
Where the power sits
In the ecosystem now forming, one line is worth translating out of the technical jargon: the registry layer is where the real power is — whoever operates the most-used registries shapes which capabilities agents find and trust[^2]. Plainly: it isn't the prettiest listing that wins, it's whoever is in the directory the AI actually consults.
At Plexa we operate exactly that: an ARD registry of local businesses for Mexico and LATAM. We publish a discoverable catalog and answer agents' discovery queries. We say it as a fact, not an announcement: it is, as far as we know, the only agentic registry focused on the region's long tail of local businesses.
Let's be honest: it's very early
None of this means "everyone already uses ARD." It's the opposite, and honesty here matters more than enthusiasm. The standard is version 0.9, emerging. On launch day itself, a census of 39 major sites — including the eleven companies in the working group — found that zero served a discoverable catalog[^2]. Zero. Real adoption, today, is near nil.
What does exist are the first pieces moving: GitHub launched an agent finder for Copilot, Hugging Face a discovery tool with semantic search, and Google said it will add native support in Gemini's agent registry in the coming months[^1]. The standard was just born, almost no one has it, and that's precisely why the first-mover window is open right now.
Because in discovery, position is captured early. The directory that already aggregated the market — that already has the clinics, restaurants, and agencies listed and verified — when agents start asking in earnest, is the one that gets cited later. Arriving once everyone's already inside is arriving late.
You don't have to understand ARD to benefit from it, just as you don't understand your email's protocol to send a message. What you can do is make sure you're inside the directory before agents start asking about your vertical in your city. Claiming your Plexa listing is that first step: it puts you in the registry that already answers.
How this article was made
This article was produced by Plexa Editor Agent.
Sources consulted: → Google Developers Blog — Agentic Resource Discovery (ARD) announcement, June 17, 2026. → Agentic Resource Discovery — the standard's open specification (agenticresourcediscovery.org).
Reviewed and published by: a human.
See the workflow → /en/agent-made
Frequently asked questions
How this article was made
This article was produced by Plexa Editor Agent.
Sources consulted:
- →Google Developers Blog — Agentic Resource Discovery (ARD) announcement, June 17, 2026. 11-company coalition; day-1 implementations (GitHub, Hugging Face, Google Gemini).
- →Agentic Resource Discovery — open specification (Apache 2.0), built on the Linux Foundation AI Catalog data model. June 18, 2026 census: 0 of 39 major sites served a discoverable catalog.
Reviewed and published by: a human.
See the workflow →Make sure you're in the directory before they ask.
Claiming your listing is the step to be inside the registry that already answers when an agent asks about your vertical in your city.
Claim your listing →