Lighthouse Agentic Browsing category
Agentic Browsing is an experimental category in Lighthouse. It checks how well a page works for AI agents: programs that browse websites and fill in forms on someone’s behalf. Unlike the other categories, it doesn’t give a score from 0 to 100. It shows how many of its checks the page passed.
Why it matters
Section titled “Why it matters”Agents read a page differently from people. They rely on the same structure that screen readers use: names for buttons and fields, a clear order of elements. A page that’s easy for an agent to use is often easier for people with assistive technology too.
The category is new, and the standards behind it are still proposals. Google describes it as a way to collect data and give hints, not as a ranking. Passing every check doesn’t guarantee that an agent can complete a task on the page.
What influences it
Section titled “What influences it”Browser: Most checks look at the page’s accessibility structure: whether buttons, links and form fields have names, and whether the roles of elements fit together. Layout shifts count as well, because an element that moves can make an agent click the wrong thing.
WordPress: The theme and form plugins produce most of the buttons and fields that an
agent would use. Other checks look for newer, optional additions: a summary for agents in a file
called llms.txt at the root of the site, and WebMCP, a proposed way for a site to describe its
actions to agents. These usually come from a plugin or custom code, if at all.
In Permo
Section titled “In Permo”Permo doesn’t record this category.
Reference values
Section titled “Reference values”Testing the category requires Chrome 150 or later, and the WebMCP checks need registration for a Chrome origin trial (source: Lighthouse agentic browsing scoring, accessed October 16, 2026).