Build Agent‑Friendly Websites: Google’s New Design Guidance for AI Agents
Audit your site for semantic HTML and stable layouts to make it agent‑friendly.
Audit your site for semantic HTML and stable layouts to make it agent‑friendly.
Summary
Google’s web.dev now advises developers to treat AI agents as a distinct audience, highlighting that complex hover states and shifting layouts break agent navigation. The guide explains that agents interpret sites via screenshots, raw HTML, and the accessibility tree, and recommends semantic HTML elements, stable layouts, and proper accessibility markup. Specific suggestions include using <button> and <a> tags instead of styled <div>s, linking <label> tags to inputs with the for attribute, and setting cursor:pointer on clickable elements. Google also introduces WebMCP, an early preview program that allows websites to register tools with input/output schemas for agent function calls. The guidance aligns with existing accessibility and performance best practices, benefiting both agents and human users.
For sites already following accessibility standards, little change is needed. For those that do not, adopting semantic HTML and stable layouts extends the business case for accessibility to AI agents that browse, compare, and transact on behalf of users. The WebMCP preview program is open for sign‑ups, and Chrome plans to discuss it at Google I/O in May.
Developers should audit their sites for semantic markup and layout stability to ensure compatibility with AI agents and to improve overall user experience.
Key changes
- Google adds agent‑friendly design guidance
- Agents use screenshots, raw HTML, and accessibility tree
- Recommendations: semantic elements, stable layouts, proper accessibility markup
- WebMCP early preview allows function calls via schemas
- Benefits extend to human users as well