If you’ve logged into Google PageSpeed Insights recently or had your web agency run a report on your site you may have noticed something new sitting alongside the familiar Performance, Accessibility, Best Practices, and SEO scores.
There’s a fifth one now. It’s called Agentic Browsing.
Most business owners in Australia have no idea it exists yet. Most agencies haven’t written about it. We’re one of the first to break it down in plain language, because what it signals about the future of websites is something every business owner should understand now, not in 12 months when everyone else is scrambling to catch up.
What Is the Agentic Browsing Score?
Google’s PageSpeed Insights is powered by an open-source tool called Lighthouse. It’s the engine that generates all those scores when you test your website. In May 2026, Google quietly added a brand-new Lighthouse category, ‘Agentic Browsing’ and it started showing up in reports almost immediately.
Here’s what makes it different from every other score:
- Performance measures how fast your site loads for human visitors.
- Accessibility measures how usable it is for people with disabilities.
- SEO measures whether search engines can crawl and index your pages.
- Agentic Browsing measures something none of the others touch: how easily an AI agent can read, understand, and interact with your website.
Rather than showing a number out of 100, it shows a pass ratio, something like 3/3 or 2/3 indicating how many of its checks your site passes.
Why Does This Exist?
Think about how you used Google five years ago. You typed a query, got a list of links, and clicked through to a website yourself.
Now think about how people are increasingly using AI tools in 2026. Someone might ask ChatGPT to “find me a reliable plumber in Brisbane” or use Google’s own AI tools to compare three local accountants. These AI systems don’t browse the web the way humans do, they visit websites on a person’s behalf, scan them programmatically, extract information, and report back.
Your website is no longer only visited by human eyes. It’s being read by machines acting for humans and Google has now started formally grading you on how well you serve them.
We’ve run PageSpeed Insights on a range of client sites since this score appeared. The results were mixed. Some sites passed all three checks without any changes. Others, including some with strong traditional SEO scores were failing two out of three. A perfect SEO score does not guarantee you pass the Agentic Browsing checks. They measure different things entirely.
What Are the Three Checks?
Google currently grades your site on three default checks in the Agentic Browsing category. Here’s what each one means in plain English.
1. Accessibility Tree (Is Your Site Structured So a Machine Can Read It?)
This checks whether your website’s underlying code gives AI agents enough context to understand what they’re looking at. Things like do your buttons have proper labels? Are your headings structured logically? Do your images have descriptive text?
This overlaps with accessibility for human users, which is one reason well-built sites often pass this check without extra effort. But sites built quickly with drag-and-drop tools, or sites that rely heavily on visual design without clean underlying code, often struggle here.
2. Cumulative Layout Shift (Does Your Page Stay Still?)
Cumulative Layout Shift (CLS) has been a Core Web Vitals metric for human visitors for a few years; it measures how much your page “jumps around” as it loads. If buttons and images shift position unexpectedly, that’s a poor experience for users.
What’s new is why this matters for AI agents: if a page keeps shifting while an agent is trying to interact with it, trying to read a form, find a phone number, or identify a call to action, it can’t complete the task reliably. Google is now counting this as an AI readiness signal too.
3. llms.txt (Have You Given AI Models a Map of Your Site?)
This is the check that a lot of Australian business website is currently failing and it’s the most interesting one.
An llms.txt file is a plain text file you publish at the root of your website (like yourwebsite.com.au/llms.txt). Its purpose is to give AI language models (like ChatGPT, Gemini, Perplexity, and Copilot) structured guidance about your business, what you do, what pages exist, and how to make sense of your content.
Think of it like a cover letter for AI. When a person visits your website, they can read around, look at images, and piece things together. When an AI agent visits your website on a person’s behalf, it benefits enormously from having a clean, plain-language summary it can reference instantly.
The good news is that if your site was built by us, you’re likely already covered on this one. All of our websites run Yoast SEO, which automatically generates an llms.txt file for your site. It’s one of the reasons we’ve used Yoast as our standard for years.
Does This Affect Your Google Rankings Right Now?
The honest answer is not directly, and we’d rather be upfront about that than oversell it.
Google has marked this category as “under development,” and it doesn’t currently feed into traditional search rankings the way Core Web Vitals do. The score shows a pass ratio rather than a weighted number, and Google itself has been careful not to call it a ranking factor.
But that framing misses the bigger point.
Google doesn’t add a formal scoring category to their most-used web performance tool, one that millions of developers and agencies check every week, without signalling a clear direction. They did exactly the same thing years ago with mobile-friendliness and page speed before those became hard ranking signals. The pattern is consistent: Google measures it before it penalises for it.
More immediately, your AI visibility is already being affected right now. When someone asks ChatGPT or Perplexity for a recommendation and your competitor’s site is easier for that AI to read, understand, and summarise than yours, they get the recommendation and you don’t.
What Does a Well-Optimised Site Actually Look Like?
Rather than sending you down a technical rabbit hole, it’s worth understanding what each of the three checks is actually looking for. That way you have a feel for whether your site is likely in good shape or whether it’s worth a conversation.
A clean, well-structured build. The accessibility tree check is essentially asking whether your website’s underlying code is logical and well-labelled. A site built properly on WordPress with a quality page builder, using correct heading structures, descriptive image labels, and properly named buttons and links, will generally pass this without any extra effort. A site that was thrown together quickly or has accumulated a lot of messy edits over the years is more likely to have issues.
A stable layout that loads predictably. The layout shift check rewards sites where elements don’t jump around as the page loads. This is often affected by things like images that don’t have defined dimensions, fonts that load late and push content around, or sliders and popups that aren’t configured carefully. It’s not something you’d notice as a visitor, but it matters to both Google and AI tools trying to interact with your page.
An llms.txt file. As covered above, if your site was built by us you already have this handled through Yoast. For sites built elsewhere or on other platforms, this is worth checking.
The honest truth is that most business owners don’t need to dig into any of this themselves. What matters is that the agency who built and maintains your site is across it. If you’d like us to review how your site is sitting against these checks, we’re happy to take a look and give you a straightforward summary of where things stand.
The Bigger Picture
There’s a useful way to think about where this fits: SEO made your site findable. AI readiness makes it usable.
For the past decade, the goal of a well-built website has been to rank in Google and convert human visitors. That’s still true. But a growing share of the traffic that drives those conversions is now being filtered through AI, whether that’s someone asking an AI assistant for a recommendation, using AI-powered search summaries, or relying on tools that browse the web on their behalf.
Google’s Agentic Browsing score is the first formal acknowledgement that this shift is real enough to measure. The businesses that start preparing now, while most people still haven’t noticed the fifth score sitting in their report, will be best positioned when it matters more.
We’re already updating our build process and working through existing client sites. If you’d like to know how your website currently scores or want us to make the changes for you, reach out and we’ll take a look.
