AI agents are adding a third audience to the web alongside people and conventional search crawlers. They read pages, answer questions and may take actions on a person’s behalf, increasing automated requests even when no human visits the site. WordPress has useful foundations for this model: feeds, structured content and the REST API already allow external software to read—and, where authorized, write to—site data.
Cloudflare Radar reported that bots accounted for about 57% of HTML requests across its network in June 2026, compared with 43% from people. That figure includes crawlers and does not identify how much traffic comes from task-performing agents, but it reflects a shift in how content is accessed. Readable HTML, semantic markup, accessible control names, descriptive image text and stable layouts can help both crawlers and agents interpret a WordPress site.
The standards and tools remain experimental. WebMCP is under development, with Chrome and Edge testing it through origin trials; Safari and Firefox do not support it. Lighthouse’s experimental Agentic Browsing checks cover agent-readable controls, layout stability, llms.txt and WebMCP, while Lighthouse 13.5 also checks for /.well-known/ai-catalog.json. These checks offer practical diagnostics, not a settled compliance standard: WordPress developers should treat agent readiness as an evolving interoperability concern rather than a fixed checklist.
Author of the original text: Miriam Schwab (Elementor)
