Google Modern Web Guidance 1.0: AI Coding Guidance for GitHub Copilot and Existing Projects

For developers using GitHub Copilot, this is a practical addition to the toolbox—not a new coding model or a guarantee of flawless websites. The project supplies targeted guidance that agents can retrieve while working. Its repository documents support for the GitHub Copilot CLI.

Better implementation choices, not just newer styling​

Google says the guides address gaps in accessibility, performance, security, and user experience. They incorporate Baseline compatibility information so recommendations can reflect a project’s browser-support requirements. User-facing guides also receive specialist accessibility review.

The important distinction is between making a page look modern and choosing appropriate platform capabilities. Google’s examples include using native HTML and CSS where an agent might otherwise reach for additional JavaScript. That does not mean every older implementation needs replacing; compatibility requirements still matter.

Installation and an existing-project scan​

Google documents this installation command, run from the project directory:

npx modern-web-guidance@latest install

Existing installations using that method can update with:

npx modern-web-guidance@latest update

These are the documented entry points, rather than a separate Windows-specific installation procedure.

The repository also provides a scan command:

npx modern-web-guidance@latest legacy-scan .

For GitHub Copilot CLI users, its alternative plugin setup is:

/plugin marketplace add GoogleChrome/modern-web-guidance
/plugin install modern-web-guidance@googlechrome

The generic installer runs an interactive wizard. During subsequent tasks, an agent can search for relevant guidance and retrieve a particular guide into its context. The repository describes guide search as local, without API keys or network calls; downloading or updating the package is a separate matter.

What the benchmark improvement actually means​

The project’s November 8 evaluation snapshot reports unguided-to-guided pass-rate improvements of 58% to 94%, 54% to 82%, and 57% to 88% across three agent configurations. Its evaluation uses simulated development tasks and Playwright checks covering source patterns, computed styles, runtime behavior, and accessibility states. These are project-reported results—not independent measurements or a forecast for every application.

Keep the browser tests​

There is an important boundary around modernization: Google says version 1.0 uses static analysis against a limited set of patterns. It can examine HTML, CSS, JavaScript, and package.json, but it is not a comprehensive application audit. Google separately recommends runtime verification with browser tooling.

The sensible adoption strategy is therefore incremental: try the guidance on a bounded task, review the proposed changes, and test the resulting application against your actual requirements. Better advice is useful. It still should not be handed the production keys without supervision.

Correction (November 15, 2026): The legacy-scan command was printed as "npx modern-web-guidance@latest legacy-scan." — the correct command is "npx modern-web-guidance@latest legacy-scan .", with a space before the dot (the directory path to scan).