What the Agentic Browsing score checks and how to pass it – Fuel Made

Last updated: July 2026

Key takeaways

Google added an Agentic Browsing category to Lighthouse 13.3 in May 2026, and it now appears in PageSpeed Insights by default
The score is a pass ratio (like 1/3 or 3/3), not a 0–100 number, and it can change as Google develops the category
The core check is your accessibility tree, the same structure screen readers use to read your site
AI traffic to US retail sites grew 393% year over year in Q1 2026, according to Adobe
Accessible, semantic code passes the audit; there's no AI-specific rebuild required

The Agentic Browsing score is a Lighthouse audit category, added in May 2026, that measures how easily AI agents can read, navigate, and complete tasks on a website. Google PageSpeed Insights reports it as a pass ratio, like 1/3, based on checks of your accessibility tree, layout stability, and llms.txt file.

It sits right next to the four scores ecommerce teams already track: Performance, Accessibility, Best Practices, and SEO. And unlike those four, it grades your site for an audience that isn't human.

What the Agentic Browsing score checks

The Agentic Browsing score checks four things: the integrity of your accessibility tree, your layout stability (Cumulative Layout Shift), the presence of an llms.txt file, and WebMCP integration. PageSpeed Insights typically reports three of these, since WebMCP audits only apply to sites that have implemented the protocol.

Check	What it tests	Effort to pass
Accessibility tree	Interactive elements have programmatic names, valid roles, and correct parent-child relationships	Moderate; this is standard accessibility work
Layout stability (CLS)	Content doesn't shift while loading, so an agent doesn't click where a button used to be	Low if your Core Web Vitals are already healthy
llms.txt	A machine-readable markdown summary of your site exists at the domain root	Low
WebMCP	Forms and site actions are annotated so agents can use them directly	High, and optional; the protocol is experimental

Why Google added a score for AI agents

Google added the score because a growing share of web traffic is agents acting on behalf of people, not people themselves. AI assistants like ChatGPT, Perplexity, and Google's own Gemini now research products, compare prices, and complete tasks across websites. Those agents need to understand your pages to do any of that.

The traffic data backs this up. Adobe Analytics, which tracks over one trillion visits to US retail sites, r eported that AI-sourced traffic grew 393% YoY in Q1 2026, following a 693% jump during the 2025 holiday season. And that traffic buys: as of March 2026, visitors arriving from AI sources converted 42% better than non-AI traffic.

The accessibility tree is the heart of the score

The accessibility tree is a simplified, structured version of your page that browsers build from your HTML. Screen readers use it to announce content to blind and low-vision users. AI agents use the same tree, because it's far more efficient to process than raw HTML or a screenshot.

Every interactive element on your site needs a programmatic name, roles and relationships need to be valid, and interactive content can't be hidden from the tree. A button that a screen reader announces as "button" with no label is a button that AI agents can't identify either.

The work that makes your store usable for a shopper with a screen reader is the same work that makes it usable for an AI agent shopping on someone's behalf. No need for a separate "AI optimization" project in this case. It's the same semantic HTML, correct ARIA labels, and machine-readable structure that accessibility-first teams (like ours) have been shipping for years.

The score is not a Google ranking factor

Google hasn't made the Agentic Browsing score a ranking factor, and the company has said that llms.txt isn't used for Search ranking or AI Overviews. Nor does the score affect your Performance, Accessibility, Best Practices, or SEO numbers. Each category is calculated independently.

Treat it as a readiness signal instead, at least for now. It’s possible that this will affect rankings someday (Core Web Vitals started as Lighthouse metrics before becoming ranking signals in 2021). Regardless of what happens later, the underlying a ccessibility and stability work already pays off in conversion, usability, and legal risk reduction today.

How to improve your Agentic Browsing score

Improving an Agentic Browsing score means fixing your accessibility tree first. Here's the order of operations:

Run PageSpeed Insights on your homepage and one revenue page, like a PDP, and note which checks fail
Fix accessibility tree failures: add missing labels to buttons, links, and form fields, and correct invalid ARIA roles
Check image markup while you're in there, since alt text failures are among the most common accessibility flags
Stabilize your layout by setting explicit dimensions on images and reserving space for late-loading content like banners and reviews
Add an llms.txt file, a plain markdown file at your domain root that summarizes your site and links to key pages
Skip WebMCP for now unless you have engineering appetite for experimental protocols

Test the pages where an agent would actually transact. A perfect homepage score means little if your product and cart pages fail.

What this means for Shopify stores

For Shopify stores, the Agentic Browsing score is a free diagnostic of how ready your theme is for AI-assisted shopping. The audit runs at the page level, so Shopify stores are scored the same way as any other platform. Theme quality is what decides the result.

Fuel Made builds and remediates Shopify themes to WCAG 2.2 AA standards, which is why the stores we work on tend to pass the accessibility tree check without agent-specific changes. This requirement has always been important; Google’s audit just turned it into a public score.

FAQ

Why is my Agentic Browsing score 1/3?

A 1/3 score often means your layout stability (CLS) check passes but your accessibility tree check and llms.txt check fail. This is a common result for ecommerce sites, since Core Web Vitals work is more widespread than accessibility work.

Is the Agentic Browsing score a ranking factor?

No. Google hasn't tied the Agentic Browsing score to Search rankings, and it has stated that llms.txt doesn't affect Search or AI Overviews. The accessibility fixes it encourages do support usability and conversion, though.

Does a good Accessibility score guarantee a good Agentic Browsing score?

No. The two categories are scored separately. Agentic Browsing filters a specific subset of accessibility audits focused on machine interaction, and it adds layout stability and llms.txt checks that the Accessibility category doesn't include.

Do I need an llms.txt file?

An llms.txt file takes about an afternoon to create, so it's a reasonable addition even though its benefit is unproven. It's a markdown file at your domain root that summarizes your site for AI models. Google has confirmed it doesn't affect Search or AI rankings.

What is WebMCP?

WebMCP is an experimental protocol that lets a website describe its actions, like adding an item to cart or submitting a form, in a format AI agents can read directly instead of inferring from page structure.

How do I check my Agentic Browsing score?

Run your URL through PageSpeed Insights at pagespeed.web.dev, where the category now appears by default. You can also run Lighthouse from Chrome DevTools, or from the command line with the latest Lighthouse CLI. Test your homepage plus the pages where a purchase happens, like product and checkout pages.

Fuel Made works with Shopify and Shopify Plus brands on accessible design, development, and ADA compliance. If you'd like help increasing your store's scores, you can reach our team here.