Lighthouse agentic browsing scoring

# Scoring Agentic Browsing Results in Lighthouse Source: https://developer.chrome.com/docs/lighthouse/agentic-browsing/scoring?hl=en The "Agentic Browsing" category assesses how well your site is designed for machine interaction through a set of specific audits. ## How the Category Score Is Calculated Unlike other Lighthouse categories, the "Agentic Browsing" category does not have a weighted average score from 0 to 100. Because web standards for AI agents are still under development, the current focus is on collecting data and providing actionable signals rather than a final ranking. Instead of a score, the report shows: - **Fractional result**: A percentage indicating how many of the website readiness checks are passed. - **"Pass" or "Fail" status**: Some audits may display errors or warnings if technical requirements (e.g., WebMCP schema validity) are not met. - **Pass count**: The category header may include a *pass rate* to help you quickly assess overall progress. ## Why Results May Vary Although the audits are specific, your results may fluctuate due to changes in how your site registers tools or responds to agentic requests. Common reasons include: - **Dynamic tool registration**: If your site registers WebMCP tools using JavaScript (the Imperative API (https://developer.chrome.com/docs/ai/webmcp/imperative-api?hl=en)), the timing of these registrations can affect whether they are detected during Lighthouse snapshotting. - **Variability in accessibility tree structure**: Significant changes in DOM size or complexity can impact the accessibility tree structure, a key metric for agent-based navigation. - **Cumulative Layout Shifts (CLS)**: Layout shifts caused by ads, images without dimensions, or injected content can move elements between the time an agent identifies them and when it tries to interact with them. ## How Audits Are Determined Lighthouse uses a set of specific signals to evaluate your page, ensuring audits are reproducible and suitable for integration into CI/CD pipelines. ### WebMCP Integration Lighthouse calls the `WebMCP` domain in the Chrome DevTools Protocol (CDP) to track tool registration events, checking both declarative tools (defined in HTML) and imperative tools (defined in JS). ### Agent-Relevant Accessibility Automated agents rely on the accessibility tree as their primary data model. Lighthouse filters a specific subset of accessibility audits that are essential for machine interaction, such as: - **Names and labels**: Ensuring every interactive element has a programmatic name. - **Tree integrity**: Verifying that roles and parent-child relationships are valid. - **Visibility**: Ensuring content is not hidden from the accessibility tree while still being interactive. ### Stability and Content Findability - **Cumulative Layout Shifts (CLS)**: Measures visual stability, which is critical for agents that depend on element positioning. - **llms.txt**: Checks for an available machine-readable summary at the domain root. ## What Developers Can Do to Improve To make your site more compatible with agentic technologies: - **Use WebMCP**: Use the WebMCP API to explicitly expose your site's logic and models to AI agents. - **Ensure a healthy a11y tree**: Prioritize semantic HTML and proper ARIA labeling, as these are your page's "machine view." - **Optimize for stability**: Minimize layout shifts to ensure agents can reliably interact with the UI without unpredictable element movement. ## Feedback If you encounter bugs or have suggestions for changes to the agentic browsing audit, please follow these steps: