WordPress.com introduces AI agents to write and publish posts automatically

The WordPress.com web hosting platform is deepening its use of artificial intelligence by introducing AI agents that can draft, edit, and publish content directly on users’ websites. The update, announced on Friday, signals a shift that could significantly reshape how content is created and managed across the internet.

With the new functionality, AI agents can handle a wide range of tasks beyond writing. They can manage comments, update and correct metadata, and organise website content through tags and categories. All of these actions are controlled through a conversational interface, in which site owners describe what they want using natural-language instructions.

This development opens the door for websites to be largely built and maintained by AI systems under human supervision. While this lowers the barrier for launching and running a website, it also raises the possibility of an increasingly populated internet by machine-generated content rather than material written by people.

WordPress, as a platform, powers more than 43% of all websites globally. Although WordPress.com represents only a portion of that total, its network still holds a significant presence, generating around 20 billion page views and attracting approximately 409 million unique visitors each month.

The introduction of AI agents builds on WordPress.com’s earlier adoption of MCP (Model Context Protocol), which was rolled out last fall. MCP is a framework that allows applications to supply contextual data to large language models. Through this system, AI tools can already connect to WordPress.com sites and access information such as content, settings, and analytics via external applications like Claude Desktop, Cursor, and Visual Studio Code.

With the latest update, AI agents can now go beyond simply reading site data. They can actively create new content, including blog posts, landing pages, and About pages, and even make structural adjustments to the website itself.

At launch, the agents will also be able to moderate and respond to comments, clean up discussions, and reorganise categories and tags across a site. They can improve search engine optimisation by updating elements such as alt text, captions, and titles. All actions performed by the AI are recorded in the platform’s Activity Log, ensuring transparency for users.

Users still retain control over the publishing process. They can prepare drafts for the AI to finalise and categorise, including adding meta descriptions, or they can let the AI generate entire posts from a prompt. By default, all AI-generated content is saved as a draft and requires user approval before publication.

Despite these safeguards, the expanded capabilities could dramatically accelerate website creation, especially when human involvement in content production is minimal.

The system is also designed to maintain visual and structural consistency. Before generating content, the AI can analyse the site’s theme and layout, allowing it to match existing design elements such as fonts, colours, spacing, and block patterns.

To activate these features, WordPress.com users can visit wordpress.com/mcp and enable the desired capabilities. They can then connect their preferred AI tools — including Claude, Cursor, ChatGPT, or other MCP-compatible applications — and begin using AI to build and manage their sites.

The introduction of these tools is likely to spark debate about the future of online content. However, some argue that AI-generated posts can also provide transparency into how these systems function and communicate. Similar experiments have already taken place, including Meta’s acquisition of Moltbook, a platform where AI agents interact with one another, and Anthropic’s trial of an AI-written blog monitored by human editors.