Google Advises Developers To Support AI Agents

Google's web.dev site published a guide titled "Build agent-friendly websites," which frames AI agents as a distinct audience alongside human visitors, per Search Engine Journal's coverage of the guidance. The guide describes three primary ways agents may interpret web content, listed in the guidance and reported by Search Engine Journal:

• •Screenshots, which let agents use vision models to identify elements visually.
• •Raw HTML, which provides the DOM structure and hierarchy.
• •A high-fidelity map supplied by the site, described as stripping visual noise to expose interactive elements.

The guidance recommends developer practices that emphasize accessibility and semantics, as reported by Search Engine Journal. Examples called out in the coverage include preferring semantic elements like <button> and <a> over styled <div> elements, keeping layouts stable across pages, linking <label> tags to inputs via the for attribute, and setting cursor: pointer on clickable elements. The guide explicitly states, "Everything we suggest to make a site 'agent-ready' also makes sites better for humans," per the published guidance.

Search Engine Journal reports that the guide links to WebMCP, a proposed web standard for agent-website interaction, and that Chrome's team describes WebMCP as an early preview program accepting developer sign-ups. The coverage summarizes WebMCP's intent as letting websites register tools with defined input/output schemas that agents can discover and call as functions.

Treating AI agents as a distinct visitor type formalizes work that overlaps heavily with existing accessibility and semantic-HTML best practices. Industry-pattern observations show that improving semantic markup, labeling, and layout stability reduces ambiguity for automated parsers and vision models, while also improving human usability.

Priorities implied by the guidance, framed generically, are to improve semantic HTML, ensure form controls are properly labeled, avoid brittle layout/hover-only interactions, and provide clear affordances for clickable elements. Adopting these patterns typically lowers maintenance friction when automated tools parse pages and simultaneously raises baseline accessibility and testability.

Observers should track WebMCP's preview program and developer sign-ups, browser-team updates from Chrome about agent APIs, and any formal specification activity around site-discoverable tools or callable I/O schemas. Wider adoption by search engines, browsers, or major agent platforms would be the next indicator of practical standards uptake.