Search Engine Journal reports that Google’s Modern Web Guidance 1.0 adds an audit for existing code and brings Firefox and Edge into the project. The free set of packaged instructions is designed to help AI coding agents write code using current browser features.
Philip Walton, a senior staff developer relations engineer at Google, announced the release in a Chrome for Developers post. Chrome first offered the guidance as an early preview at Google I/O in May. Google says teams can use it with their coding tools while developing new projects or reviewing existing code.
Modern Web Guidance Audits Existing Projects
The version 1.0 audit examines a project’s HTML, CSS and JavaScript files, along with its package.json. It identifies areas where developers can safely update code, then guides the agent through those changes. The first iteration uses static analysis and covers a limited number of code patterns, according to Google.
Google says the library now targets “hundreds of verified AI knowledge gaps” across its guides. Domain tech leads review and approve each guide. Guides for user-facing features also receive accessibility review from experts who use assistive technologies daily.
The project’s governance document, uploaded to GitHub on Nov. 9, outlines six owner seats. Two belong to Chrome, while Microsoft Edge and Mozilla each have one. Independent web platform experts hold two seats; one of those seats is open, and Apple has no seat.
Teams can install the guidance in a project directory with the command described in Google’s announcement, or find it in plugin marketplaces for most popular coding agents. Google’s post also links to a public dashboard where Anthropic, OpenAI and Google coding agents run the same tasks nightly, with and without the guidance.
Tests Show Different Results With Guidance
A browser checks each finished app’s DOM, styles and source code. In results for nightly runs completed Nov. 8, guided runs passed 88% of automated checks, compared with 56% for unguided runs. The dashboard combined each agent’s results across 199 to 215 tasks. Its accessibility review is limited and does not claim to be a full WCAG audit.
Google also highlighted testing by Japan’s creator platform note inc. Two agents completed six interface tasks across 180 test runs on clone versions of production code. On 19 browser checks, one agent’s score rose from 60.8% to 76.5% and then 94.0% as the guidance moved from absent, to installed, to named in prompts. The second rose from 75.8% to 86.2% and then 90.0% under those conditions.
Those results sit alongside practical browser work such as Chrome’s soft navigation metrics, which track in-app page changes, and WordPress design updates such as expanded responsive design controls. Site teams also manage everyday interface features, including the new Jetpack form and video tools.
Google’s source describes the guidance as a way to bring current browser practices into agent-written code, including existing projects. It plans to expand the audit process and suggests pairing the guidance with Chrome DevTools for agents. The published tests show gains in their stated checks, while Google’s dashboard notes that its accessibility review is limited.
