Guide: How To Check If Your Website Is Ready For AI Agents

Guide: How To Check If Your Website Is Ready For AI Agents

AI agents are shifting how people interact with websites. Instead of browsing directly, users increasingly send AI agents to navigate sites, fill out forms, and complete tasks on their behalf. Websites that agents can't parse, navigate, or interact with risk being bypassed entirely.

Google has released a new Lighthouse audit category called "Agentic Browsing" that evaluates how well your site supports this shift. It checks whether AI agents can identify interactive elements, understand your site's structure, and use your tools — and it runs in under five minutes with no technical skills required.

This isn't about search ranking. It's about whether an AI agent sent by a potential customer can actually do anything useful on your site. Right now most websites fail at least some of these checks (including Google's own documentation pages), so the goal is establishing your baseline and knowing which fixes to prioritize.

What You'll Need

Browser

• Chrome Canary (free) — Google's beta version of Chrome. The standard version of Chrome does not include the Agentic Browsing category yet.

Other

• Your website URL

Time: Under 5 minutes per page

Step 1: Install Chrome Canary

Go to google.com/chrome/canary and download Chrome Canary. Install it like any other browser — it runs separately from your regular Chrome and does not replace it.

Open Chrome Canary and navigate to your website.

One important note: do not use Chrome Canary as your everyday browser. It is a beta version and can be unstable. Use it specifically for running this report.

Step 2: Run the Agentic Browsing Audit

With your website loaded in Chrome Canary, open DevTools by right-clicking anywhere on the page and selecting Inspect (or press Ctrl+Shift+I on Windows / Cmd+Option+I on Mac).

In the DevTools panel, look at the tabs along the top — Elements, Console, Sources, and others. Click the >> arrows if you can't see all of them, then find and click Lighthouse.

In the Lighthouse panel you'll see checkboxes for different audit categories: Performance, Accessibility, Best Practices, SEO. Look for the new one: Agentic Browsing. Check that box. You can also check Accessibility if you want, since both are relevant to agent readiness.

Choose whether to test on Mobile or Desktop. Start with Desktop, and test both if you have time.

Click Analyze page load and wait 30–60 seconds. Your page may briefly resize or reload during the process — this is normal.

Step 3: Interpret Your Results

The report does not give you a score out of 100 like a traditional Lighthouse performance audit. Instead, it shows a ratio of how many agentic readiness checks your site passes or fails.

You'll see two groups: passed audits (no action needed) and failed audits (areas where agents may struggle with your site). The report evaluates three main areas:

Accessibility Tree

This is the most immediately actionable area for most websites. The accessibility tree is the structure originally built for screen readers, but AI agents now rely on it too. It tells agents where the buttons are, what elements are important, and what is clickable.

If the report says your accessibility tree is "not well formed," agents will struggle to navigate your site. Share the specific failures with your developer — fixes typically involve adding proper ARIA labels, fixing area roles, and ensuring interactive elements are correctly identified. These are standard accessibility improvements that any web developer will understand.

WebMCP

WebMCP is a proposed standard that lets you teach AI agents how to use the tools and forms on your website. It comes in two types: declarative (simple code wrapped around a form telling agents what it does and how to use it) and imperative (a more advanced setup allowing agents to interact back and forth with your site).

If the report shows WebMCP items as "Not Applicable," that simply means your page doesn't have WebMCP set up — which is the case for the vast majority of websites right now. You only need to investigate this if you have forms, calculators, booking tools, or other interactive functionality that you specifically want AI agents to use on behalf of their users. If your site is primarily informational or product listings, this is not urgent.

LLMs.txt

This is a file (similar to robots.txt) that gives AI agents specific instructions about what they're allowed to do on your site and where to find important information, written in markdown format.

An important distinction: Google has said you do not need an LLMs.txt file for search ranking. That's true, but this report isn't about search — it's about agents using your website directly. You likely don't need this file unless you have tools, APIs, or resources specifically designed for AI agent use. When you start building WebMCP-enabled tools, that's when an LLMs.txt file becomes valuable.

Step 4: Prioritize Your Fixes

Not everything in the report needs immediate attention. Here's how to triage:

Act on now: Accessibility tree failures. These affect both human users on screen readers and AI agents, so fixing them improves your site for everyone. Share the specific audit failures with your developer.

Keep on your radar: WebMCP. If you have interactive tools on your site — quote calculators, booking forms, product configurators, search tools — start thinking about how to make these agent-accessible. If your site is primarily content or product listings, this isn't urgent but will become important.

Park for later: LLMs.txt. Unless you're already building agent-facing tools, you don't need this file today. When you do, you or your developer can create it using the documentation at llmstxt.org.

Quick Reference

What This Means For Your Business

The agentic web is the shift from people browsing websites directly to people sending AI agents to accomplish tasks on their behalf. This audit is an awareness exercise — most sites won't pass every check, and that's expected at this stage. The goal is knowing your baseline so you can make informed decisions as agent readiness becomes a competitive factor.

The full SOP with additional context on implementation priorities is available here

• OpenAI updates its EU ads policy to serve ads only to users who explicitly opt in, relying on consent rather than legitimate interest under GDPR, as ChatGPT ads launch in the UK and it eyes $102B ad revenue by 2030.

• OpenAI plans a desktop superapp merging ChatGPT, Codex and its browser to streamline the user experience and refocus on engineering and business customers, with Fidji Simo and Greg Brockman overseeing the revamp.

• Google confirms a small US test of healthcare ads in AI Mode for English queries, with PMax, AI Max, Shopping and broad match eligible, though the first iteration bars creatives using pinning or text disclaimers.

• Social commerce grows four times faster than traditional e-commerce, projected to reach $8.5T by 2030, as creator partnerships drive revenue and brands running influencer plus affiliate programs see a 46% sales uplift.

• Anthropic suspends Fable 5 and Mythos 5 after US national security authorities flag a jailbreak method, cutting off Pro, Max and Enterprise preview users, with no timeline for return as its Pentagon dispute continues.

• Bots overtake human web traffic for the first time at 57.4%, as AI agents synthesize content without sending readers or revenue to creators, threatening the ad and subscription economics underpinning the internet.

• OpenAI partners with LiveRamp to pipe conversion data into its ads API, the first independent ad tech integration, letting select US advertisers tie ChatGPT ad exposure to real-world purchases as it chases performance budgets.

• Pinterest launches Amazon Storefront linking, letting creators connect their storefront so affiliate links apply automatically when they tag eligible Amazon products, cutting manual steps from inspiration to earnings.

📦 Langfuse - Trace, evaluate and debug your AI agents and LLM apps in production.

🎶 Suno - Turn a text prompt into a full, editable song with vocals.

🎙️ Murf AI - AI voiceover platform with preset voices, dubbing and a script assistant.

🖥️ Beautiful.ai - AI presentation maker with auto-designed, on-brand slides.

📧 Apollo.io - Find, enrich and sequence outreach from a 275M+ contact database.+

😍It was great, all I needed

🙂Good, not great

🙄It sucked