On November 10, 2026, Google's Chrome team released Modern Web Guidance 1.0, the stable version of a set of guides that AI coding agents consult to write current, cross-browser HTML, CSS and JavaScript. It can now also review existing code and suggest upgrades. It matters to anyone who builds or maintains a website with AI help.
What we know
- What it is: according to the Chrome for Developers blog, it is a set of "skills" that gives coding agents web platform expertise, best practices and browser compatibility data. It was previewed at Google I/O in May.
- Modernization: the headline 1.0 feature is a workflow that scans a project's HTML, CSS, JavaScript and package.json for safe upgrade opportunities and guides the agent through them. Google notes that this first release uses static analysis and only checks a limited set of patterns.
- More guides, more review: the announcement describes a guide database aimed at "hundreds of verified AI knowledge gaps," with sign-off from a tech lead in each area and an accessibility review for user-facing guides.
- Compatibility: the guides build on Baseline data, which tracks which web features work across the major browsers, so recommendations can match the browsers a project needs to support.
- Firefox and Edge: they are joining Chrome to support the project, improve cross-browser compatibility and contribute.
- Where it runs: install it with `npx modern-web-guidance@latest install` in the project folder, or through the plugin marketplaces of Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build.
- Results: Google publishes an evaluation dashboard and company case studies. In one, note inc. says its modern web check pass rate rose from 61% to 94%. These are the companies' own numbers, with no independent measurement.
What changes and what doesn't
- The agent's starting point changes. Google points to verified AI knowledge gaps: techniques agents don't know or get wrong. With these guides, the agent knows which current technique to use and where it works.
- It doesn't replace testing. Modernization is limited and the analysis is static: it doesn't check how your site looks or loads on a real phone.
- It doesn't boost SEO by itself. Cleaner, modern code can help speed and accessibility, but Google doesn't present the tool as a ranking factor.
- It's open. Content is licensed under Creative Commons Attribution 4.0 and code under Apache 2.0.
How to tell if it affects you
- Check how your site gets built. If you or your developer use Claude Code, Cursor, Copilot or another agent on the list, this is useful today. If your site is only edited from the WordPress or Drupal dashboard, nothing changes for you.
- Try it on a copy. Install it on a separate branch or a staging site, never straight on the live site, and review every change the agent proposes before accepting it.
- Limit what it touches. Point it at your theme, templates and your own scripts. WordPress or Drupal core, third-party plugins and modules, and the dependencies folder are updated separately: if the agent edits them, the next update will wipe those changes.
- Set your target browsers. Check in your analytics which browsers and versions your visitors use and tell the agent, so it doesn't reach for features part of your audience doesn't have yet.
- Measure before and after. Run PageSpeed Insights on your key pages before applying changes and again afterwards. If something gets worse, roll it back.
- If you don't code, ask. Ask whoever maintains your site: do you use an AI agent for code? With which guides? Do you test changes before deploying them?
Related: Agent Skills Moves to the AAIF: What It Means for You
Sources
Updates: if Google expands the modernization workflow or publishes new evaluation numbers, they will be added here with a link.