Key points
- Modern Web Guidance 1.0 is free, installs with one npx command, and adds an audit of existing HTML, CSS, JavaScript and package.json files.
- Firefox and Edge join the project, whose governance document lists six owner seats: two Chrome, one Edge, one Mozilla, two independent (one open). Apple has none.
- On Google's dashboard, guided agent runs passed 88% of automated checks versus 56% unguided, but guided runs were generally slower and used more tokens.
- MetaMask's review bot flagged 102 issues across 57 pull requests in five weeks; engineers fully resolved 43 of them.
Why it matters for marketers: Teams using AI coding agents get a free way to steer output toward current browser features, but the evidence is mostly Google's own, so human accessibility and field testing still matter.
Google has released version 1.0 of Modern Web Guidance, a free set of packaged guidelines that AI coding agents consult while writing web code. The aim is to get agents to use current browser capabilities instead of outdated patterns. Philip Walton, a senior staff developer relations engineer at Google, announced it in a Chrome for Developers post. It first appeared as a preview at Google I/O in May.
Version 1.0 adds an audit that scans a project's HTML, CSS, JavaScript and package.json to find places where code can be safely updated, then guides the agent through the changes. Google says this first audit uses static analysis and covers a limited set of patterns. Each guide is now reviewed by a domain tech lead, and user-facing guides get an accessibility review by experts who use assistive technology daily.
The published results are encouraging but mostly come from Google and customers it highlighted. On Google's nightly dashboard, guided runs passed 88% of automated checks and unguided runs 56%, though guided runs were mostly slower and used more tokens. Japanese creator platform note inc. saw both tested agents improve, and MetaMask's review bot flagged 102 issues in five weeks. Accessibility consultant Adrian Roselli, testing the earlier preview, found output that did not meet WCAG.
For content and web teams, the practical step is to try it on a staging project and keep field data and manual accessibility testing in the review loop. Google also suggests pairing it with Chrome DevTools for agents so the agent can run and test its own code in a browser. Watch for the planned expansion of the audit and for whether the open independent governance seat gets filled.
Sources
This brief is our own summary. Read the full reporting at the source.