Google's Modern Web Guidance 1.0 — The Free Toolkit That Teaches Your AI Coding Agent to Write Modern Web Code

Your AI coding agent can build a working website in minutes — but the code it writes is often years out of date. Google's Modern Web Guidance 1.0, released November 10, is a free open-source skill pack that teaches agents the modern web, and Firefox and Edge just joined in. Here's what it does, what the numbers really mean, and why it matters for anyone building sites with AI.

Google's Modern Web Guidance 1.0 — The Free Toolkit That Teaches Your AI Coding Agent to Write Modern Web Code
Google's Modern Web Guidance 1.0 — The Free Toolkit That Teaches Your AI Coding Agent to Write Modern Web Code — Technical Architecture & Overview

Table of Contents

I have a confession about the AI-generated code I have been shipping lately. It works — the page loads, the button does the thing, the layout looks right. And then, when I read what the agent wrote, I find the same problem: the code is old. Not broken old. Just 2019 old. A pile of divs and JavaScript doing a job the browser has done natively for years.

That is exactly the problem Google just tried to fix. On November 10, 2026, Google's Chrome team announced Modern Web Guidance 1.0 — a free, open-source set of skills you install into the coding agent you already use, so it stops writing web code the old way. It is not a new AI model or a new coding agent — in Google's own framing, a crash course in the modern web for your agent. And the part that made me sit up is who joined: Firefox and Edge.

The problem: your AI agent writes 2019 code

AI coding agents are trained on the internet's code — billions of lines of it, most of it written years ago. When you ask an agent to build a modal or a dropdown, it reaches for the patterns it has seen most often — and those are the old ones: a div with a click handler, a JavaScript library where a native browser feature would do, a custom solution for something the platform solved years ago.

The result is code that works and is quietly worse than it should be. It is heavier than it needs to be, because it ships JavaScript the browser does not need. It is less accessible, because a hand-rolled dropdown does not behave like a real one for a screen reader or a keyboard user. Philip Walton, the Google engineer who announced the release, put it plainly: models can usually produce something that works, but a web expert can almost always make it faster, more accessible or more secure. I have felt this in my own WordPress and Elementor work — the agent delivers something functional, then I spend an hour fixing focus states and mobile behaviour. Its training data taught it the web of five years ago.

What Modern Web Guidance 1.0 actually is

Modern Web Guidance is a library of expert-written guides that your coding agent consults while it works — less like a tool, more like a senior developer sitting next to the agent, whispering the right answer at the right moment. When the agent is about to build something, the guidance tells it which modern browser feature fits the job, how to use it, and what the browser support looks like.

Version 1.0 ships with 203 use-case guides covering 250 modern web features: 140 on CSS and layout, 74 on JavaScript and APIs, and 36 on HTML and DOM — every one expert-reviewed, and the guidance adapts to your Baseline browser target. It is free and open — software under the Apache 2.0 licence, guide content under CC-BY 4.0, on GitHub at github.com/GoogleChrome/modern-web-guidance. Install it with one command, npx modern-web-guidance@latest install, or from the plugin marketplaces of Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build. Google previewed it at Google I/O in May 2026; six months later it is stable. It is not a model or an agent. It is a knowledge layer.

What is new in 1.0: a health check for your old code

The genuinely new thing in version 1.0 is the modernization workflow — a scanner for existing code. Until now, the guidance only helped with new code. Version 1.0 can look at a codebase you already have and find the outdated patterns in it. A follow-up Chrome post on November 11 breaks it into four steps: scan, verify, report a plan, and modernize. The scanner reads your HTML, CSS, JavaScript and package.json, flags where a modern approach would be better, and guides your agent through the changes.

There are honest limits. The scanner covers only about a fifth of the guide catalogue, and it works by static analysis — reading your code without running it. There is no automatic fix mode: it finds the opportunities and plans the work, but it does not rewrite your code by itself. That is a deliberate choice, and the right one. Even so, this is the part I am most interested in, because most of us maintain sites that already exist — client sites, old themes, plugins written years ago. A tool that audits that code and points at where a modern browser feature would replace a hundred lines of JavaScript is worth more to me than one that writes slightly better new code.

The cross-browser part is the real news

Here is the sentence in the announcement that matters most: Firefox and Edge have joined the project. Modern Web Guidance is now a cross-browser effort, with Chrome, Microsoft Edge and Mozilla Firefox contributing and maintaining it, alongside MDN. The governance makes this concrete: six owner seats — Chrome two, Edge one, Mozilla one, and two independent web-platform experts.

That structure matters, because the obvious fear with a Google-led project is that it becomes a vehicle for Chrome-only features. The governance, and the presence of Mozilla and Microsoft in it, is the answer to that fear. Philip Walton said the guides document each feature with a browser support table for anything not Baseline Widely available, usually with a cross-browser fallback. One glaring absence, though: Apple is not in it. Safari is the third major browser engine, and every iPhone user reaches your page through it. The guides still carry browser support data — but the project is cross-browser in the sense of "Chrome, Edge and Firefox," not "every browser your users actually use."

The numbers — and why I am skeptical of them

Google's announcement comes with striking numbers. In its nightly evaluation runs, coding agents using the guidance passed 88 percent of automated checks, against 56 percent without it. A follow-up post put it as roughly 50 percent without and around 90 percent with. Case studies back it up: the Japanese platform note inc. ran two coding agents through six interface tasks across 180 test runs, and its check pass rates rose from 61 percent to 94 percent. MetaMask now runs the guidance as part of an automatic review of every pull request to its browser extension.

Real results — but vendor-reported numbers from Google's own test suite, not independently reproduced results. The 88-versus-56 figure comes from nightly runs up to November 8 using the preview-era guidance, so it is evidence from the preview period, not a measurement of version 1.0 itself. Treat them as a signal, not a verdict. What convinces me more is the shape of the evidence: multiple companies reporting the same direction of improvement, and a public nightly dashboard anyone can check.

What this means for Elementor and WordPress developers

Why am I writing about this on a site about Elementor and WordPress? Because this is where our work is heading, fast. AI agents have already moved into every page builder — Elementor has its own MCP server now, and agents can build real Elementor pages from plain-English prompts. They are building client sites today, and the quality of their code is becoming the quality of the web itself.

That is what makes Modern Web Guidance relevant to us. When an agent builds a custom widget for an Elementor page or modernizes an old theme, the guidance is the difference between code that works and code that works well — accessible, performant, maintainable, built on features the browser already has. For a freelancer or an agency, that difference shows up in what clients actually notice: pages that load faster, forms that work with a keyboard, layouts that do not break on a phone. And the modernization workflow has the most immediate value for WordPress work, because most WordPress sites are not new — years of accumulated themes, plugins and custom code.

How to try it in five minutes

From your project directory, run npx modern-web-guidance@latest install. The wizard asks which agent you use and wires the guidance in. If you prefer marketplaces, it is available as a plugin for Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build. Once installed, you do not need to do anything special — the agent consults the guides as it works. But several case studies found the guidance worked better when the agent was explicitly told to use it. A line like "follow Modern Web Guidance for web platform choices" in your prompt is cheap insurance.

For the modernization workflow, point it at an existing project and let it scan. Read the plan before you let it change anything — the scanner finds opportunities, it does not apply fixes on its own, and that is a feature, not a limitation. Google also recommends pairing the guidance with Chrome DevTools for agents, so the agent can verify its own code inside the browser — the honest answer to the guidance's biggest limitation: the guides can improve the choices an agent makes, but they cannot prove the result works. Treat agent changes as reviewable pull request drafts, and test the result in a real browser before it goes near production.

Conclusion

Modern Web Guidance 1.0 is not the most exciting announcement of the year. It is not a new model, it does not generate anything by itself, and it will not write your next website for you. What it is, is something quieter and possibly more important: an admission from the people who build the web platform that AI agents have a knowledge problem — and a practical, free, open-source attempt to fix it. A library of 203 expert-reviewed guides, a scanner that audits the code you already have, and cross-browser governance that includes Firefox, Edge and MDN. It is a young tool with honest limits — it cannot prove the result works, and the scanner covers only a fifth of the guides. Even so, the web is about to be built, in large part, by agents. Anything that makes their code more modern, more accessible and more performant is worth five minutes to install.

FAQ

What is Modern Web Guidance 1.0?
A free, open-source set of skills from Google's Chrome team that embeds web platform expertise into AI coding agents. It is not a model or an agent — it is a library of 203 expert-reviewed guides covering 250 modern web features, plus a scanner for existing code. Released as stable 1.0 on November 10, 2026, after a preview at Google I/O in May.

How do I install Modern Web Guidance?
Run npx modern-web-guidance@latest install from your project directory — it launches an interactive wizard. It is also available through the plugin marketplaces of Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build.

Does Modern Web Guidance replace my coding agent?
No. It is a knowledge layer alongside the agent you already use. The agent still writes the code; the guidance tells it which modern browser features to reach for, and what the browser support looks like.

What is the modernization workflow?
A scanner, new in version 1.0, that analyzes your HTML, CSS, JavaScript and package.json to find safe modernization opportunities, then guides your agent through four steps: scan, verify, report a plan, and modernize. It covers about 20 percent of the catalogue and applies no fixes automatically.

Is Modern Web Guidance really cross-browser?
Mostly. Firefox and Edge have joined, and governance is shared across Chrome, Edge, Mozilla and two independent experts, with MDN involved. Apple and Safari are not part of it.

Should I trust the 88% versus 56% evaluation numbers?
Treat them as a signal, not a verdict. They come from Google's own nightly runs using preview-era guidance, not independent tests of version 1.0. The convincing evidence is the pattern: multiple companies reporting improvements, and a public dashboard anyone can check.

Looking for more WordPress & Elementor tutorials?

Explore our complete directory of 53+ web performance, theme development, and Elementor architecture guides.

View Topic Sitemap XML Feed