The core contributors behind WordPress’ AI initiate have been building off the very successful AI building blocks that James LePage put in motion just one year ago. The work is great and I’m very optimistic about what we’re seeing in the broader ecosystem.
Meanwhile, I have been thinking about it from a WordPress product and marketing perspective. Big picture, we need to make sure the WordPress brand is more closely aligned with the future of web development by leaning deeper into outside-in agentic site management. The building blocks are there, but are we showing a compelling user experience?
An ecosystem approach
WordPress is open source, infinitely accessible and extensible, so AI enablement should be an easy story to tell. We’re not doing a great job of telling it. The gap is mostly being filled by plugins and hosts (often on top of core plumbing like the Abilities API and MCP Adaptor) or users are being pushed outside of WordPress to different frameworks and static site generators.
Knowledge travels the internet in different ways than in the past. Ecosystem diversity and community-sourced informational content is still important, but the WordPress project needs a clear, opinionated path as more websites start their journey not at a hosting company’s front door, but in their AI agent.
We also have to think holistically about the three important layers of our ecosystem:
- WordPress’ open source core and the WordPress brand
- Hosting companies like WordPress.com, Pressable, Siteground, Kinsta, etc
- Plugin ecosystems like Jetpack, Elementor, AwesomeMotive, and WooCommerce.
Even harder – we need everything to work together as a cohesive unit, offering a single entry point and experience for users. And we must do this while also ensuring there’s room for each layer to provide additional value, monetize, or build loyalty where appropriate.
The building blocks were a successful start. Now that we’re further in, how should we prioritize and market those building blocks? Where can/should we dive head-first into the trends, and where should we keep our wait-and-see approach?
I’m going to give my thoughts and- to be clear- I’m thinking of this from the perspective of a developer advocate, a marketer, and most importantly, a user who just wants to show you cool things with WordPress and WooCommerce. This is more go-to-market thought exercise than a technical roadmap, because the technical pieces are good. And I might be wrong.
Grow the composable architecture
The Abilities API (in core since 6.9) has the strongest signal for demand. We’re already seeing adoption across the ecosystem and we should continue promoting it as the core plumbing for agentic workflows. Yet WordPress core as a content management system has no content management Abilities. It’d be like shipping a WP REST API with no endpoints, or WP CLI with no commands.
A new Core Abilities plugin should be released as a standalone feature plugin with a suite of core read-write abilities. Those Abilities should eventually migrate to core itself, but a feature plugin makes sense for quicker testing and iteration. I should be able to fetch a set of patterns, design a page, post a blog post, and update my header and menu, all via Abilities.
There are discussions about adding Core Abilities into the MCP Adaptor or AI Experiments plugin, but I disagree with this approach. AI Experiments should not be a dependency for anyone using MCP. Basic content abilities are a fundamental primitive beyond either of those plugins because they help us create a composable agent-first management layer. The sooner we get them in core, the better we can promote the WordPress MCP as a useful tool.
MCP is good enough
Will the MCP protocol be around in a decade? Will another protocol take its place? I’m not sure, but I’m not worried about it, thanks to our approach. It’s around now and there’s enough demand that I think it’s worth promoting it is a first class tool, just like Laravel, Shopify, Vercel, Figma, Spacefast, and many other web development tools have. MCP itself doesn’t need to be in core, but it should be canonical and much more discoverable.
For now the package is bundled in major plugins like Woo and Elementor, but I’m glad to see it’s going to become a separate plugin. (We just need to drop “Adaptor” from the name. It’s cleaner.)
The only issue is that as a plugin, it doesn’t really do anything for core users without Core Abilities being available. Installing it is a non-event. It should include a clear onboarding screen for setting up Claude/Codex/etc, and possibly some basic governance, like disabling Abilities that it exposes. We have all these pieces in the various AI plugins, just not in a single, user-friendly way.
AI and MCP for the product ecosystem
Personally, I’d consider AI tools in the dashboard that require an API key a lower priority for core than solving the external agent access problem (without requiring full REST or SSH access). Building the UI and plumbing for AI interactions might’ve been harder a year ago, but this is something an agent can spin up pretty quickly whether core provides a framework or not.
Consider what core can offer inside of WordPress right now – generate an excerpt or a featured image? Write a post for you? I mean, that’s useful, but the more impactful work will most likely happen in your external bot that has access to Google Drive, Linear, Github, spreadsheets, etc. Plus bulk processing across multiple entities, bulk site updates and management, bespoke theme design, custom blocks and admin screens, etc. Lets prove it first where people are actually using AI before we add sparkle buttons to core.
The Abilities API, on the other hand, offer so much more value because of the compounding effect we get when every plugin uses it. It aggregates functionality into a single layer.
If we survey the landscape, the AI Experiments plugin has 50k active installs but only 9 reviews, so it’s unclear how much of these installs are ‘active’ in the way we’d consider it. I’d love to see more data on this because my theory is that ‘outside-in’ AI approaches are winning right now.
Other interesting AI plugin stats
Other AI plugins are hard to compare, because install numbers might be even less organic. Elementor’s Angie – 100k installs, only 18 reviews. Siteground’s agent – 1m installs but overwhelmingly negative reviews (due to how they rolled it out) and only 14 positive reviews.
The WPVibe MCP plugin, to compare, has 40k installs in a fraction of the time and 77 positive reviews already.
The only comparable AI (but also MCP) plugin with a strong dotorg appearance is Jordy Meow’s AI Engine – 90k installs, almost 900 positive reviews. It’s a hard one to compare because it also includes a chatbot-building tool and premium offerings, like a dashboard AI agent.
Even with a core set of Abilities, there’s still plenty of room for innovation and additional value from hosts and plugins. For AI features in the dashboard, this is where I’d let good, existing admin experience plugins like Jetpack and Elementor step up and fill the gap, like they already are. This is that extra layer of value add that other tools can provide.
Hosts and plugins might also layer on additional value with custom MCP connections: governance, audit trails, analytics and additional integrations, multi-site management via a single connection, and other ideas we haven’t even seen yet. They can more securely offer advanced abilities like plugin and theme management, rollbacks, and more. Those are premium offerings with relevance and outside the scope of core content management.
Plugins with MCPs can also register branded Plugins for Claude and ChatGPT. A user might go searching specifically for WordPress, Jetpack, or WooCommerce, and each could present their own branded plugin, and those connections will still surface all registered Abilities.
Create a marketing home for anything AI/MCP on WordPress.org
Ultimately, WordPress.org needs to have an opinion on AI. A basic Google search shows no core information about how to use AI with WordPress, and the top URLs are taken by third-party plugins that may or may not work for everyone. Much of the information that is available from Core is for building AI tools for WordPress, nothing about managing your WordPress site with AI.
There’s a few things users might want to do with WordPress, and an agent needs to know the best way to do all of it:
- Want to connect your site to your existing agent? Install MCP or browse popular plugins.
- Want to empower AI features in the dashboard? Install AI Experiments or browse popular plugins.
- Want to build a site from scratch? We need a one-line install of Playground CLI, Agent Skills, etc
Then the developer site can focus on these developer tools around building with the AI blocks or even contributing to WordPress with AI via the Plugins MCP, Trac MCP, Contributor Toolkit, etc.
There’s a few steps to telling this story, beyond just a clarified approach to the building blocks themselves:
- Targeted landing pages for humans and agents that tell the story: /ai, /mcp, /try, etc
- Include copy/paste prompts and AGENTS.md files available
- Promotion in the WordPress dashboard for installing canonical abilities and MCP
- Clear CLI tooling and SDKs that don’t require multiple tools strung together
- Centralized developer resources for AI on https://developer.wordpress.org/
Tools like Spacefast are doing a great job at this AI-first marketing approach.
To be clear: The Core AI contributors are doing a phenomenal job, and this post isn’t complaining about any single project, but treating it like a top-to-bottom go-to-market strategy for WordPress AI, and prioritizing based on our users’ needs, not technical tooling.
Last thoughts
It’s hard to predict AI. A few weeks ago I was really worried about agents struggling with the complexities of block markup. As the frontier evolves, I’m less worried about that.
Now I’m more worried that an agent isn’t going to recommend WordPress in the first place because it doesn’t know how much it can do with it, because the tools aren’t clearly accessible, and because WordPress.org isn’t telling anyone about it.
What do you think?
Leave a Reply