Our answer, after living in all of these tools, is more interesting than a yes or a no. WordPress is unquestionably behind on a few things that matter. It’s also quietly doing some of the most important work in the space right now, and the people writing its obituary may end up eating a little humble pie.

Where the native AI actually is

For a long time “AI in WordPress” meant a grab-bag of third-party plugins bolted on to write blog intros. That’s changing at the core level. The official WordPress AI plugin now handles the everyday content work directly inside the block editor: title suggestions, alt text for accessibility, excerpts and meta descriptions, summarizing, translating and resizing content, plus image generation and editing right where you’re writing. It even does editorial review block by block, comment moderation with toxicity and sentiment detection, and tag and category classification.

More importantly, this isn’t a one-off plugin. It’s a reference implementation on top of new foundations that shipped in WordPress 7.0. Core now includes an AI Client, an Abilities API that lets any plugin expose its own AI-powered functions, and a Connectors Hub where you authenticate providers like Anthropic, Google or OpenAI in a few clicks. There’s no vendor lock-in baked in; you bring your own model. WordPress isn’t trying to be the AI. It’s building the plumbing so the whole ecosystem can plug in. That’s a very WordPress way to approach the problem, and it tends to age well.

The honest caveat: the plugin is still officially experimental and it’s block-editor only. But the direction is unmistakable.

Going past copy and images with MCP

The native tools are great for the light stuff: words, images, metadata. The heavier lifting is where things get exciting, and where a plugin like WPVibe changes the equation.

WPVibe turns a self-hosted WordPress site into an MCP server, which is the connection standard that lets AI agents actually do things rather than just chat. Once it’s set up through a clean OAuth flow, with no application passwords pasted into a chat window, you can point Claude, ChatGPT, Cursor, Windsurf, or any MCP-compatible agent at your site and have it work directly. That means creating and publishing posts, pages and custom post types through the REST API, uploading media, running a set of allowlisted WP-CLI-style commands via native PHP, and even editing your theme through a sandboxed draft-preview-publish workflow with PHP syntax validation before anything touches production.

In plain terms: your preferred AI agent can inspect your site, find the right files, make structural changes, and ship them safely. That’s a categorical step beyond “write me a paragraph.” It’s the difference between an assistant that suggests copy and one that can rebuild a template.

The “just leave” narrative doesn’t hold up

Here’s where the popular story gets thin. The pitch for the new agent-based builders is essentially WordPress is old, this is new, so move. But “new” and “better for your specific job” aren’t the same thing, and the trade you’re being asked to make often isn’t spelled out.

WordPress is genuinely open source. You own the code, the database, and the CMS structure, and you can model your content however your business actually works. Most of the shiny alternatives are proprietary and, more to the point, structured to nudge you up the pricing ladder every time your needs grow. That’s not a knock on the companies. It’s just the model.

And the limits show up in oddly specific places. Webflow, as capable as it is, caps you at 20 CMS collections on its CMS plan and 40 on Business, with a hard ceiling of 30 custom fields per collection no matter what you pay. If your content model is rich, you feel those walls fast. Framer is a beautiful tool, but try doing something as routine as fluid typography with font-size: clamp() and you’ll find yourself writing a code override just to force a class onto an element so you can target it with your own CSS. These aren’t dealbreakers. They’re the small, recurring reminders that you’re a guest in someone else’s system, decorating within the lines they’ve drawn.

WordPress’s answer to “can I build this?” is almost always yes, because you can edit the CMS itself. Whatever you can imagine, you can model. That freedom is the whole point, and it’s exactly what a locked, tiered platform can’t give you.

Where WordPress is honestly still behind

We’re not here to wave pom-poms. There are real gaps.

Out of the box, a dynamic WordPress site rendering pages on every request will rarely match the raw page-speed of a static-first platform that ships pre-built pages to a CDN. You can close most of that gap with good caching and architecture, but it’s work the newer platforms hand you for free. Native accessibility, while improving, still leans heavily on the theme and the person building it rather than being guaranteed by the platform. And for all the progress above, the AI story is younger and more fragmented than what a purpose-built modern tool ships on day one.

If you only weighed today’s defaults, the case to leave would look stronger.

What would actually move the needle

The encouraging part is that none of these gaps are structural. They’re roadmap items, and WordPress has the largest development ecosystem on the web pushing on them.

Imagine core shipping a genuinely static-first or hybrid rendering path so dynamic sites get CDN-speed delivery without a stack of caching plugins. Imagine a full site editor that’s fast and predictable enough that building a layout feels less like negotiating with the interface. Imagine theme development modernized around reusable, well-documented patterns, and the one we’d most like to see, the ability to define a custom style guide and apply those styles by class, the way developers actually think, rather than only by semantic element. That single shift would make WordPress design systems feel first-class instead of improvised.

None of that is science fiction. Given who’s working on it and how fast 7.0 moved, a lot of it feels like a matter of when.

The Lightwave take

We don’t have a favorite horse in this race. Lightwave Creative’s job is to pick the best tool for the specific job in front of us, and sometimes that genuinely is Webflow or Framer. They’re excellent at what they’re good at, and we reach for them when the fit is right.

But the reflex to write WordPress off feels premature. It owns an enormous share of the web, it’s the only major player that’s truly open, and it’s evolving faster right now than it has in years. The AI foundations are in core, agents can already operate on real sites through MCP, and the roadmap points straight at the gaps critics keep naming. Betting against a platform with that much momentum and that big an ecosystem has burned people before. We’d be a little careful about ordering the humble pie in advance.