Over the last few months, I’ve been thinking about how WordPress fits into the AI era of the internet. There’s still a strong need for a robust CMS, especially in ecommerce, government, enterprise, publishing, and more. But when I think of the WordPress community, it includes the vast ecosystem of small SMBs, marketing sites, and agencies. There’s been a shift away from WordPress towards static or small headless sites, and tools like Astro and Next.js have become popular with agencies and individuals using AI coding tools.
Last week I had the pleasure of joining this conversation with a smart group of WordPress legends. The panel included Matt Medeiros (The WP Minute, Gravity Forms), Mike McAlister (Ollie), Alex Panagas (Scalemath), and Andy Peatling (Miles, ex-Automattic). You should watch the full recording here, and then keep reading for some thoughts I’ve had since then:
In the podcast, we focus on two important questions: should WordPress continue to compete in this ‘simple website’ space, and if so, what needs to change about the project, the marketing, and the software itself?
A lot of the recent core AI work has been focused on making it easier to work on sites that are already running WordPress: the MCP Adapter has been released in the plugin directory, and third-party tools like Mike’s Ollie and Andy’s Miles are giving us robust capabilities for site design and management. That work is still important when WordPress is the chosen solution, but Andy specifically has been been focused on a different question: how do we ensure that WordPress is the recommended choice by agents for new site builds?
Andy’s been running some benchmarks to track this. As for solving it, there are a lot of potential ideas that some of us have discussed, but each one gives you a lot to consider:
Do we simplify the block theme architecture, making it easier for agents to assemble a frontend the way they can with other frameworks? You can check out tools like Block Runner from Human Made or liberate.sh from Automattic that convert static sites to WordPress, but neither approach makes WordPress the default markup for an agent. There’s also promise in headless WordPress, separating content from design entirely, which is why I’ve long advocated for content modeling in core. That said, I think we lose a lot when we abandon or try to hack our way around the block editor.
Do we drive down the price by going static? Static sites are much cheaper to host, especially when you add the fact that you can skip extras like security and backup plugins. Tools like Spacefast and Simply Static are pushing boundaries there, but you have serious economic tradeoffs when you cut off most of the plugin ecosystem and start operating at much lower price points than traditional WordPress hosting.
Do we make WordPress simpler to spin up, removing the database and focusing on an entirely code-based workflow? Spinning up and managing HTML content via a database is some heavy friction for an agent. Like many others, I’ve experimented with a git-based workflow that stored post content as markdown and used an ephemeral Playground to still give you the WordPress content editing experience. But again, there are some serious tradeoffs to that approach when you push WordPress that far from it’s typical plugin-based configuration.
Do we do better marketing on WordPress.org of the existing tools and AI capabilities, something I recently wrote about? That’s probably tables stakes. Some things could be quick, like better, indexable documentation, more video content, and A/B testing some proper landing pages. Some will take more time, like building a deep catalog of content to influence future models. (Luckily for us, Nick Diego is back, and it looks like he is on the case.)
Every option has tradeoffs, but so does ignoring the current environment. Change WordPress too much, and we lose the backwards compatibility and the vast plugin ecosystem. Ignore the evolving landscape and risk obsolescence. These are hard, interesting problems, and as bullish as I am about the new AI tools in WordPress, I actually think it’s going to be what we remove from WordPress that’s even more important. Finding ways to remove complexity and abstractions that deter agents will be just as critical as adding sparkle buttons and an MCP server.
Overall, I’m excited to have these conversations, and to see just how fast some of my colleagues can move on their ideas with AI in their toolkit.
Watch the full conversation on YouTube and let me know your thoughts.
Leave a Reply