WordPress.com, a prominent web hosting platform, has announced a significant integration of AI agents, a move poised to fundamentally alter how websites are created and managed. This development allows AI agents to draft, edit, and publish content on customer websites, alongside managing comments, updating and fixing metadata, and organizing content with tags and categories. Users can control these functionalities through a natural language interface, simply explaining their desired actions.

This new paradigm means websites could be almost entirely conceptualized and maintained by AI agents guided by human oversight, significantly lowering the barrier to entry for website setup and ongoing maintenance. However, it also raises the prospect of a web increasingly populated by machine-generated content rather than human-authored material. While WordPress, as a publishing platform, powers over 43% of all websites globally, WordPress.com, its hosted version, constitutes a smaller yet substantial segment, drawing 20 billion page views and 409 million unique visitors monthly.

These expanded AI capabilities build upon the introduction of Model Context Protocol (MCP) support by WordPress.com last fall. MCP is a standard designed to provide context to large language models (LLMs), enabling AI assistants to connect with the platform. This prior integration allowed customers to gain visibility into their site's content, settings, and analytics through their preferred AI applications, such as Claude Desktop, Cursor, or VS Code. Now, AI agents are empowered not only to read site content but also to actively create new posts, landing pages, and About pages, as well as implement structural modifications.

Upon launch, the AI agents offer a diverse set of management tools. They can approve, reply to, and clean up comments; create, rename, and restructure categories and tags across the site; and correct alt text, captions, and titles to enhance the site's search engine optimization (SEO). All these changes are meticulously tracked within the site’s Activity Log, ensuring transparency and accountability.

Customers have flexible options for content creation: they can author drafts for their AI agent to publish, tag, and categorize, complete with a meta description. Alternatively, they can instruct their AI agent to generate a post or page from scratch by simply describing the desired content. To maintain user control, WordPress.com emphasizes that all changes necessitate user approval, and any posts written by AI are saved as drafts by default. Despite these safeguards, the enhanced capabilities are expected to drastically accelerate website creation, especially in scenarios where human content generation is minimal.

Furthermore, the company highlights that the AI agent can analyze the site’s existing theme and design before commencing content creation. This ensures that the generated content aligns with the site's aesthetic, utilizing consistent colors, fonts, spacing, and block patterns. To activate these new features, WordPress.com customers need to visit wordpress.com/mcp and toggle on the desired capabilities, then connect their chosen MCP-enabled AI client, such as Claude, Cursor, or ChatGPT, to begin. While concerns about the nature of web content are anticipated, the company suggests that AI-authored posts can offer valuable insights into how these advanced models compose and interact. This concept has precedents, with Meta acquiring Moltbook, a social network where AI agents posted and interacted, and Anthropic's experiments with AI-written blogs under human supervision.