WordPress Meets AI: Why 'No CMS' Is Suddenly on the Table

TL;DR

• WordPress now powers **42.4%** of the web — the first dip below 43% since 2022
• The "None" category (sites with no CMS) rose from 28.6% to 29% — the first increase in roughly a decade
• AI tools like Claude Code & Codex are making it faster & cheaper than ever to build sites without a CMS
• Some developers are leaving WordPress entirely; others are using those same AI tools to go deeper with it
• Chris Lema, CEO of Motivations AI, puts it plainly: WordPress is still the right solution for 75–85% of websites

According to W3Techs, WordPress now powers 42.4% of all websites. This is the first time it's dipped below 43% since 2022. That alone is mildly interesting. What's genuinely surprising is the "None" category: sites running no detectable content management system at all, which climbed from 28.6% to 29%. It's the first time that number has moved upward in approximately a decade.

This isn't a WordPress-is-dying story. It's an AI story; it's one every developer, freelancer & agency owner who builds on the web should be paying attention to right now — because the same AI tools are producing completely different outcomes depending on who's picking them up.

What the Numbers Are Actually Telling Us

Before we get to the human drama, let's look at the broader CMS landscape. The shift over the past decade is striking:

Source: W3Techs CMS historical data

For a long time, the story was simple: the "None" category was steadily shrinking as CMSs became more accessible & WordPress was hoovering up most of those converts. Shopify, Wix & Squarespace each gained 0.1% since January — continuing their slow, steady rise. But Joomla & Drupal are in freefall & the "None" category just reversed course.

The "None" reversal is the tell. It's not that WordPress is suddenly failing — it's that the floor just dropped out of the barrier to building a site without one.

The AI-Powered Migration Stories

The clearest signal came from an unexpected source: Joost de Valk, the founder of Yoast SEO. If you've worked with WordPress for more than five minutes, you know who he is. He recently migrated his personal blog off WordPress & onto Astro, hosted on Cloudflare Pages, then published a post titled *"Do you need a CMS?"*

His answer: for most sites, increasingly no.

"For twenty years, 'I want a website' meant 'I need a CMS.' WordPress, Joomla, Drupal: the conversation was always about which one. That framing is outdated. People never wanted a CMS. They want a website."

The post got 706 likes, 128 reposts & 158 comments on X. The replies were a cross-section of the internet: developers nodding, WordPress loyalists pushing back, AI evangelists declaring the CMS era finished.

De Valk wasn't alone. Developers across the ecosystem have been quietly telling the same story in different ways:

• Pauline Bakhtiari, community leader at Vercel, used Claude Code to migrate her WordPress blog in an afternoon — all 367 posts converted to Markdown, images optimised, deployed to Vercel. "It feels like I'm back in my hackathon days where the only limit is your imagination," she wrote.
• Tim Leland, creator of URL shortener t.ly, moved his site from a decade of WordPress to Astro in a single session using OpenAI's Codex, with AI handling content extraction, URL preservation & theme recreation.
• Chukwuma Onyeije, the physician-developer behind Doctors Who Code, spent a morning losing a methodical fight with a plugin that kept writing the wrong URL into its configuration. By early afternoon he was rebuilding from scratch in Astro. All 89 posts, migrated by evening.

The most nuanced take came from Chris Lema, CEO of Motivations AI & a longtime WordCamp speaker. He migrated his personal site off WordPress to Astro — with Payload CMS as his content layer. His reasoning: he wants AI agents to be first-class participants in his content workflow, prompting in Claude, publishing to his site, social & newsletter without leaving the AI interface.

"I'm rebuilding because I want to play. WordPress was designed for a different job."

I'll admit I'm a little biased here — this blog runs on Payload CMS too, so Lema's migration destination resonates. There's something genuinely exciting about building a content layer that AI can read, edit & reason about natively.

Why AI & Flat Files Are a Natural Fit

Here's the underlying dynamic driving all of this: AI tools work best with simple, flat, plain-text files.

A Markdown file in a Git repository is exactly that — plain text an AI agent can read, edit, reason about & commit without needing to understand anything underneath it. WordPress, on the other hand, is layered: PHP, MySQL, plugins, themes, block editor, each one a layer of abstraction built for human navigation. AI *can* work with it — plenty of developers are proving it — but it's navigating complexity that was designed for people, not machines.

De Valk frames the logical endpoint of this as a question worth sitting with:

"If you can tell a chatbot 'update the opening hours on my contact page' or 'publish a post about our spring menu' & it edits the file, commits & deploys — what's the admin panel for?"

It's a genuine question. Not rhetorical. For a certain class of simple, text-heavy websites, the honest answer is starting to become "maybe not that much."

Earlier this month, a Reddit thread titled "Bye bye WordPress" hit the front page of r/ClaudeCode. The poster had converted two client sites to Astro & React using Claude Code, deployed both to Vercel for free & declared they wouldn't touch WordPress again for new client projects. The post reached 468 upvotes & 296 comments.

The most upvoted comment — at 228 votes — was a reality check: "Keep in mind that with a custom solution you're responsible for maintaining the code going forward."

Plenty of Developers Are Doing the Opposite

Here's where the story gets more interesting & more honest.

The same Reddit thread that spawned all those "bye bye WordPress" takes was also full of developers going the other direction entirely — using the same AI tools to go deeper with WordPress, not away from it.

Redditor yamebe described how his agency was using Claude Code to "supercharge" their WordPress workflow: theme scaffolding, security scans, update management, deployment pipelines, ACF setups, custom post types. "CC takes care of all the tedious thingies now," he wrote.

Another developer, bigtakeoff, gave Claude Code SSH access to his WordPress server & was managing content, configuration & deployment entirely through natural language. He called it a *"game changer."*

Redditor martinparets described building sites using ACF, JSON & MCP integrations at "absolute lightspeed" — with clients who still got the familiar CMS interface they expected, no retraining required. "I've heard 'WordPress is dead' for a couple decades now. Our agency can't get rid of it & we still have demand for it because it's literally still the best CMS for our clients."

Vikas Singhal, founder of InstaWP, published a direct response to the Reddit thread titled "'Bye Bye WordPress' Is the Wrong Takeaway." His argument: yes, WordPress is overkill for a five-page brochure site & it always has been. "The only reason agencies used it for those projects was because building something custom took too long. That's no longer true." But the moment a site needs e-commerce, content scheduling, role-based editing, email marketing integration, or anything meaningfully dynamic, the vibe-coded static approach breaks down fast.

Matt Mullenweg offered his own take, comparing the enthusiasm for static sites to "'This is the year of Linux desktop!'" — a joke about something that's been almost ready for mainstream users for 30 years & has never quite arrived. His point: static sites are endlessly exciting to developers & largely beside the point for everyone else.

So What Does This Actually Mean For You?

The honest answer is: it depends on what you're actually building.

• If you're building a simple, content-light site — a personal blog, a portfolio, a five-page brochure — the case for reaching for WordPress by default is genuinely weaker than it used to be. AI tools have made it fast & cheap to build something static, deploy it for pennies & get a perfect performance score. If you're comfortable maintaining it, it's worth considering.
  
• If you're building anything dynamic — e-commerce, member portals, booking systems, complex content scheduling, role-based access — WordPress or a purpose-built CMS still wins. The scaffolding WordPress provides (draft states, image handling, URL structures, user management) represents thousands of well-made decisions accumulated over 20 years. You don't miss it until you have to rebuild it.
  
• If you want AI to be a first-class participant in your content workflow — like Lema — a headless CMS approach is worth exploring seriously. Flat files & Git-based content make AI collaboration genuinely smoother. That's not theoretical; it's what this site is built on.

Chris Lema, who has probably thought about this more carefully than most, puts a useful number on it: WordPress is the right solution for 75–85% of people who need a website. His own migration, he acknowledges, is a "0.001% use case." That's not a dismissal of the migration trend — it's a useful calibration for deciding whether you're actually in that 0.001%.

The Real Story: Same Tools, Different Conclusions

Nobody really had "None" on their bingo card. For years, the threat to WordPress was supposed to come from Shopify, Wix & Squarespace — well-funded platforms with big marketing budgets. Those platforms are growing, but slowly. The actual disruption turned out to be developers rebuilding sites in an afternoon & deciding they didn't need an admin panel at all.

But the more interesting pattern — the one that might actually matter more long-term — is the developers who picked up the same AI tools & went deeper with WordPress, not away from it. Faster workflows, AI-assisted theme builds, natural language server management. The platform isn't going anywhere.

What's changed is the default assumption. "I want a website" no longer automatically means "I need a CMS." That's new. What developers do with that shift is still playing out. A single number in a CMS market share report doesn't normally make waves. But this one certainly did.