ENDA

Google Lighthouse Now Checks Whether AI Agents Can Find Your Business

2026-09-29| Updated 2026-10-04
By Morten Siert Eriksen

Agentic Discoverability just received another significant sign of support. Lighthouse version 13.5 now checks your site for an ARD file, next to the other audits for SEO, PageSpeed Insights and accessibility.

While it seems like a small, experimental check, it's also the first time a mainstream Google tool has picked up on ARD as part of website quality. Hopefully this will help more sites and businesses have their agentic tools and resources found by users and customers.

Google Lighthouse Now Checks Whether AI Agents Can Find Your Business

Topics

What changed in Lighthouse 13.5

Lighthouse 13.5.0 was released on 18 September 2026. Its release notes list a new audit, "ard-schema: add Agent Resource Discovery audit", and group it with the existing llms.txt check under agent discovery. The check sits in the Agentic Browsing category, which Lighthouse added back in May 2026. Google's describes the category as experimental for now. It gives no 0–100 score, only a ratio of checks passed, because the standards for the agentic web are still emerging.

What the ARD check actually tests

The check does two things: it looks for your ARD file, and it validates the file it finds. After reviewing the Lighthouse code (and testing the audit myself), it looks in this order:
  1. An Agentmap: line in your robots.txt
  2. A link tag in your page head with rel="ai-catalog"
  3. An HTTP Link header with rel="ai-catalog"
  4. The default location, /.well-known/ai-catalog.json
The audit check against the official ARD standard and has four outcomes:
  • Pass — the file loads and follows the standard.
  • Pass with warnings — minor issues, no errors.
  • Fail — the file has errors, or the address it points to won't load.
  • Not applicable — Lighthouse didn't find a file at all.

Why your result might say "Not applicable" even if you did it right

The ARD standard moved on shortly before Lighthouse shipped. ARD version 0.91, published on 26 August 2026, made /.well-known/ard.json the one standard location, with rel="ard" as the link type. Lighthouse 13.5 still only looks for the older ai-catalog names.

So a site that follows the current standard exactly, and publishes only ard.json, gets "Not applicable". The gap is logged in Lighthouse issue #17251, which also reports that some valid files are marked as errors.

The practical fix for now is to publish at both addresses. That's what we do on fugentic.ai.

Is this a Google ranking factor?

No — at least, Google hasn't said it is. The check sits in the experimental Agentic Browsing category, not in SEO, and nothing in the release notes ties it to Search, AI Overviews or Gemini.

That isn't the interesting part, though. Google's own tooling team now measures whether an AI agent can look up what a site offers, next to speed and accessibility audits. The question is shifting from "can AI read my page?" to "can an AI assistant find what I do, and use it?"

While this might not be a ranking factor for SEO, we might be looking at indicators for a new systems like agent finders, or other component for an Agentic Discoverability eco-system.

Passing the ARD check is not the finish line

With the widely available ARD audit sitting in Lighthouse, you are a step closer to having customers and users find you when they use AI Agents. It’s equally important to present your products and services as agentic tools and resources, so AI Agents can interact with it.

Customers are soon expecting to interact with brands and businesses through AI, not only to find them. This is the big difference from simply being mentioned in the AI answers.

The ARD standard allows you to customize the description of your products and services, and even let customers and users interact with your custom customer service agent, product fit agent, sales agent and so on. Eventually closing the gap between being found and completing a purchase, a reservation and other actions in a conventional way.

What to do now?

  1. Try Lighthouse 13.5 in Chrome DevTools and open the Agentic Browsing section.
  2. Check your ARD file with our free ARD validator, which already reads the current ard.json location.
  3. Publish your file at /.well-known/ard.json, and at /.well-known/ai-catalog.json until Lighthouse catches up.
  4. Describe what you actually do, in your customers' words.
  5. You can do all of this with ease in the fugentic platform, and also get your first sales MCP server for free.

Frequently asked questions

What is the ARD audit in Lighthouse?

A check, added in Lighthouse 13.5, that looks for an ARD file on your site and validates it against the official ARD standard.

Why does Lighthouse say my ARD check is not applicable?

Lighthouse didn't find a file at the addresses it checks. If you publish only ard.json, that's expected for now. Also publish the ARD spec at /.well-known/ai-catalog.json

Does the Lighthouse ARD check affect Google rankings?

Google hasn't said so. It sits in the experimental Agentic Browsing category, not SEO.

What is the difference between llms.txt and ARD?

llms.txt helps an AI read the content on your site. An ARD file tells an AI assistant what tools and services you offer, and how to reach them and interact with them.

Do I need both ard.json and ai-catalog.json?

The standard only requires ard.json. Publishing both covers the Lighthouse 13.5 audit too.

Background: Google's announcement of the ARD standard, and our own explainer, What is Agentic Resource Discovery (ARD)?
Return to overview
Why Google Lighthouse Now Checks for ARD