Google's Agentic Resource Discovery: What It Means for Your SEO
Google, Microsoft and Hugging Face have released the Agentic Resource Discovery (ARD) specification — a new standard that lets AI agents find, evaluate and use your website's resources autonomously. If your content isn't structured for agent discovery, you're about to become invisible to the fastest-growing search channel.
What Is Agentic Resource Discovery?
Agentic Resource Discovery (ARD) is a draft specification backed by Google, Microsoft and Hugging Face that defines how AI agents — autonomous software programmes that browse, compare and transact on behalf of users — discover what your website offers. Think of it as robots.txt for a post-search world: instead of telling crawlers which pages to index, ARD tells agents which resources, APIs and capabilities your site exposes.
The spec centres on a machine-readable file (commonly agents.json or an ai-catalog.json) hosted at a well-known URL. Agents read this file to understand what actions they can perform on your site — from reading articles to booking appointments, comparing products or pulling structured data into an answer.
Google's Search team announced ARD alongside its I/O 2026 updates, where it confirmed that AI agents now account for a measurable share of search-initiated traffic. The signal is clear: your website needs to be legible not just to humans and traditional crawlers, but to autonomous agents acting on a user's behalf.
Why ARD Matters More Than You Think
Most marketers optimise for two audiences: human readers and search-engine crawlers. ARD introduces a third — and it's growing fast.
According to Google's own Search blog, AI-powered features like AI Overviews now reach over a billion users monthly. Agentic commerce — where AI agents compare, negotiate and purchase without the user ever visiting a website — is projected to handle a significant share of routine online transactions within the next two to three years, according to McKinsey research.
Here's the uncomfortable truth: if an AI agent can't discover your content, you don't exist in that transaction. It won't click through ten blue links. It won't browse your homepage. It queries a structured catalogue and moves on.
How ARD Connects to GEO and Traditional SEO
GEO — Generative Engine Optimisation — is how you get cited by AI tools like ChatGPT and Perplexity. ARD extends this principle into the agentic layer. Where GEO focuses on being quoted in AI-generated answers, ARD focuses on being actionable by AI agents.
Traditional SEO, GEO and ARD aren't competing strategies. They're layers of the same visibility stack:
- SEO gets you ranked in search engine results pages (SERPs).
- GEO gets you cited in AI-generated summaries and answers.
- ARD readiness gets you discovered and used by autonomous agents that act on a user's intent without the user ever seeing a search result.
Neglecting any one layer means losing a growing share of qualified traffic. The marketers who treat all three as a unified programme will compound their visibility advantage.
What the ARD Specification Actually Requires
The specification is still in draft, but the core components are already clear from the published documentation by Google, Hugging Face and early implementers:
- A well-known discovery file — typically
/.well-known/agents.jsonor/ai-catalog.json— that lists your site's available resources, capabilities and endpoints in a structured format. - Structured resource descriptions — each resource entry includes a name, description, input/output schema and authentication requirements. This is how agents decide whether your site can fulfil a user's request.
- Permission and authentication declarations — you control which agents can access which resources, with what credentials and under what terms.
- Human-readable documentation links — agents can follow these to understand context, which means your existing content (blog posts, guides, documentation) directly supports agent discovery.
Early adopters like Suganthan Mohan have already shipped ARD on live sites and reported that the implementation itself is straightforward — the harder part is auditing your content and services to define what you actually want agents to discover.
Five Steps to Prepare Your Website for Agent Discovery
You don't need to wait for the spec to finalise. Here's what you can do now to get ahead:
1. Audit your content for machine readability. Structured data (JSON-LD), clean semantic HTML and well-defined headings aren't just SEO hygiene — they're the foundation agents rely on. If your pages lack structured data, start there.
2. Define your resource catalogue. List every valuable action or data set your site offers: articles, product specs, booking endpoints, calculators, tools. This becomes the basis for your agents.json file.
3. Implement or plan your discovery file. Even a minimal agents.json that declares your content resources signals to early agent crawlers that your site is agent-ready. Google's developer documentation provides a starting framework.
4. Strengthen your GEO foundations. Agent discovery and generative engine optimisation share the same underlying requirement: content that is structured, authoritative and directly answers specific questions. Every GEO improvement you make today pays dividends across the agentic layer too.
5. Monitor agent traffic. Check your server logs for agent-specific user-agent strings. Google has indicated that agentic crawlers identify themselves differently from Googlebot. Understanding this traffic now gives you a baseline to measure future growth.
The Visibility Cost of Doing Nothing
Ignoring ARD won't cause an immediate traffic drop. But consider the trajectory: AI Overviews already affect click-through rates on informational queries. Agentic commerce is reshaping how users discover and purchase services. Voice assistants, autonomous research tools and AI-powered comparison engines all rely on structured, agent-readable data.
The brands that invest in agent discoverability now will own the compounding advantage — just as the brands that invested early in SEO a decade ago still dominate organic traffic today. Content we structure for agents in 2026 will drive qualified leads well into 2028 and beyond.
One client we worked with on GEO readiness saw a 300% uplift in AI-cited traffic within 12 months. ARD readiness is the logical next step — and the earlier you move, the wider your moat.
What This Means for Your Content Strategy
ARD doesn't replace your existing content programme. It elevates it. Your blog posts, guides, product pages and case studies are the raw material that agents catalogue and serve to users. But they need to be structured correctly, described explicitly and exposed through a machine-readable interface.
This is the shift: content quality alone is no longer sufficient. You also need content infrastructure — the metadata, schemas and discovery files that make your quality visible to machines acting autonomously on behalf of humans.
The marketers who recognise this shift and act on it will gain a measurable edge in traffic, leads and conversions. The ones who wait will wonder why their high-quality content stopped generating results.