When the WordPress AI Team was set up in 2025, the goal was to create a shared foundation for integrating artificial intelligence into the CMS without being tied to a single provider or model.
In 2026 this direction has become much more concrete: WordPress is not just adding visible “AI” features to the editor, it is building infrastructure that lets plugins, workflows and agents understand what a site can do and interact with it in a controlled way.
The Abilities API is the central piece
The Abilities API, introduced on the server side in WordPress 6.9, lets you register a site’s capabilities in a structured, self-documenting way.
In practice, instead of scattered functions that are hard to discover, WordPress can expose “abilities” with inputs, outputs, permissions and descriptions. This makes them easier to use from automations, interfaces, AI tools and agents.
WordPress 7.0 brings the AI Client into core
With WordPress 7.0, a provider-agnostic AI Client arrived in core: plugins and features can talk to generative models without having to implement completely different integrations every time.
The release also introduced a central screen for managing Connectors, that is, connections to external providers and services, and a client-side version of the Abilities API for working in the browser too.
Where MCP comes in
One of the most interesting projects is the MCP Adapter. Its job is to translate WordPress abilities into the Model Context Protocol format, so that compatible assistants and agents can discover them and use them as tools.
This means a WordPress site can much more easily become part of external AI workflows, without building a proprietary integration for every assistant.
What is still experimental
Not everything the AI team develops automatically ends up in core. The WordPress 7.2 roadmap keeps several features in the AI plugin, precisely to test adoption and usefulness before a possible stable integration.
- more abilities, including future write capabilities;
- distribution and updates of the MCP Adapter;
- easier MCP activation from the dashboard;
- identity and delegation for agents;
- knowledge and guidelines that can be reused across the site;
- embeddings and semantic search;
- streaming of AI responses.
What changes for developers
The most interesting part is not learning a new “AI” plugin, but understanding the new abstractions.
- Abilities API: how to describe capabilities, inputs, outputs and permissions;
- Connectors: how to manage external services and providers;
- AI Client: how to use different models without being tied to a single integration;
- MCP: how to make WordPress usable by external agents and assistants;
- permissions and auditing: how to stop an agent from doing more than it should.
For anyone building plugins, integrations or custom systems, these skills may become far more useful than simply “knowing how to generate text with AI”.
What changes for companies and teams
For a company, the potential value lies in connecting the site’s content, data and operations to smarter workflows.
- assisted content generation or review;
- editorial support and classification;
- automations between WordPress and external services;
- semantic search across your content;
- assistants that read or update data according to defined permissions;
- workflows spanning CMS, CRM, e-commerce and other systems.
So the direction is less “let’s add a chatbot to the site” and more making WordPress a platform that intelligent software can query and act on.
Will AI replace WordPress?
The project is heading in the opposite direction: WordPress is trying to become infrastructure that is easier to use for AI tools and agents too.
This does not remove the need for architecture, development, security or integrations. If anything, it makes it more important to define data, permissions, APIs and processes properly.
What is worth learning now
For a WordPress developer who wants to stay competitive, today I would focus mainly on:
- the Abilities API;
- the REST API and permissions;
- MCP and tool calling;
- React and Gutenberg development;
- integration with external services;
- the AI Client and provider abstraction;
- security of agentic workflows.
It is a direction that also fits well with web development in general: APIs, automations and integrations matter more and more compared with knowing a single platform.
If you want to explore content designed for AI systems as well, read how to write useful content for users and AI. For integration or custom development projects, see my web development services.