WordPress Was Built for a Different Internet. Can It Adapt to the One AI Is Creating?

I ran across a LinkedIn post that made me stop for a second because the basic argument sounded almost too obvious once somebody said it out loud.

WordPress was built for a different internet.

Not a worse internet, just a different one. WordPress first showed up in 2003, when publishing something online still mostly meant a human sitting at a computer, logging into an admin panel, writing or editing something, clicking a few buttons, and hitting publish. That was the job it was built to make easier, and considering how much of the web eventually ended up running on it, I think we can safely say it did a pretty damn good job.

The post I was reading was asking a different question, though. What happens when the person operating the website is not always a person anymore?

That is where I started thinking she might really have a point.

We are moving toward an internet where somebody might tell an AI agent to update a product, rewrite a page, pull sales information, schedule an article, check performance, change a setting, or manage content without ever opening the traditional dashboard themselves. The software is still doing the work, but the thing asking it to do the work is increasingly another piece of software.

WordPress was not originally designed around that idea because, obviously, nobody in 2003 was sitting around worrying about whether ChatGPT could discover a site's available tools.

That alone does not make WordPress outdated. It would be pretty ridiculous to judge twenty-three-year-old software for not predicting an AI workflow that barely existed a few years ago. The interesting question is whether something with that much history can change fast enough when the way we interact with software starts changing underneath it.

Forty percent of the web changes the calculation.

WordPress still powers an enormous amount of the internet. W3Techs currently puts it at 40.1% of all websites and 58.6% of websites where the content management system is known. So this is not some forgotten platform hanging around because three people refuse to upgrade their blogs. WordPress is still the giant in the room.

That size is probably both its biggest advantage and one of its biggest problems.

WordPress has plugins for damn near everything, hosting everywhere, developers who know it, businesses that have depended on it for years, and millions of sites full of content that nobody wants to rebuild for fun. If you already have a WordPress site that ranks, sells products, takes bookings, collects leads, and does what the business needs, “the internet is changing” is not automatically a good enough reason to tear the whole thing down.

Anyone who has ever handled a website migration knows why. Changing systems can mean moving years of content, recreating functionality, preserving URLs, mapping redirects, testing everything again, and watching Search Console like it personally owes you an explanation. Google has an entire migration guide because getting those moves wrong can absolutely cause indexing and traffic problems.

That kind of inertia is real, but the LinkedIn post made another point I liked: migration risk is a reason to migrate carefully when you need to, not necessarily a reason to stay on the same architecture forever.

Then I looked at what WordPress is actually doing.

So I started digging into whether WordPress was actually doing anything about this agent problem, and that is where the story got a lot more interesting.

I expected to find people bolting AI tools onto WordPress from the outside. What I found instead was WordPress building the plumbing for this stuff inside the platform itself.

WordPress has already been moving away from being something that can only be operated through wp-admin for years. The REST API infrastructure started landing in Core back in WordPress 4.4, with the main content endpoints arriving in 4.7. Application Passwords came later in WordPress 5.6 so outside applications could authenticate without somebody handing over their actual account password. In other words, machines interacting with WordPress is not some brand-new idea.

What is new is how explicitly WordPress is starting to think about agents.

In May 2025, WordPress announced an official AI Team specifically to coordinate AI work across the ecosystem. Then WordPress 6.9 introduced the Abilities API, which gives WordPress, plugins, and themes a standardized way to describe what they can do using defined inputs, outputs, permissions, and machine-readable schemas. The WordPress documentation specifically calls out external systems such as AI agents as a use case.

That sounds technical because it is, but the basic idea is pretty simple.

Instead of an AI system having to guess how a random plugin works, the plugin can effectively say, “Here are the things I know how to do, here is the information you need to give me, here is what I will return, and here is who is allowed to run it.”

That is a very different way of thinking about software than building everything around a human clicking through menus.

Then I found the part that made the timing of that LinkedIn post almost perfect.

On October 7, three days ago, the WordPress AI Team released its MCP Adapter as a canonical plugin in the official WordPress Plugin Directory. MCP, or Model Context Protocol, gives AI assistants a standard way to discover and interact with tools and data. WordPress's adapter connects that protocol to the Abilities API so an AI agent can discover things a WordPress site knows how to do and, when permissions allow it, actually do them.

So while I was reading somebody asking whether WordPress belongs in an agent-driven internet, WordPress was basically down the street installing the plumbing for agents.

That changed the way I was looking at the whole argument.

The thing keeping WordPress alive is also the thing making this harder.

I do think there is something true in saying WordPress was built for a different internet. You can still see that history everywhere in the platform. Even Gutenberg has to work around the expectations of the older WordPress ecosystem. The block editor works with structured block objects while you are editing, but those blocks are serialized back into post_content because WordPress has decades of tools and themes expecting content to live there in a compatible format.

That backwards compatibility is one of the reasons WordPress has survived as long as it has. It is also baggage that newer systems simply do not carry.

If somebody builds a new content platform today, they can start with structured content, APIs, agents, permissions, automation, and machine-readable schemas as assumptions from day one. WordPress has to introduce those ideas without blowing up millions of existing websites that were built under a completely different set of assumptions.

That is a much harder job.

It also makes me think the future of WordPress probably is not going to come down to whether it is “old.” Plenty of old technologies survive because they keep adapting, and plenty of new technologies disappear because nobody actually needs them.

The question I am more interested in now is whether WordPress can change without breaking itself.

The question I am more interested in now is whether WordPress can make this new agent layer feel native instead of feeling like another thing bolted onto twenty years of previous decisions.

The Abilities API and MCP Adapter make me think the people working on WordPress understand that problem. They are not pretending the way people use software is going to stay the same forever.

That does not automatically mean WordPress wins the next version of the internet either.

Forty percent of the web is a hell of a head start, but it can also make turning the ship slow. WordPress has to modernize while keeping old themes, plugins, content, APIs, developers, businesses, and workflows alive. Every new platform trying to compete with it gets to start with a much cleaner sheet of paper.

I went into this thinking I might end up writing about why WordPress is getting left behind.

I came out of it thinking the more interesting story is whether WordPress can pull off one of the hardest tricks in technology: changing what it is without breaking the reasons people still use it.

Where I landed after all of this.

Right now, I would not bet against it.

But for the first time in a long time, I also would not assume its position is permanent.

And I think that is what makes the next few years of WordPress a lot more interesting than another argument about whether people like using wp-admin.

Sources

All articles