Introduction: the headline that keeps finding me
Every few weeks now, somebody sends me the same chart with the same worried message: WordPress is dying, should we move? I build WordPress sites for a living, mostly with Elementor, so I understand why the question lands hard.
So let me say the honest thing first: the decline is real, and widely misunderstood. WordPress is not collapsing. It is losing a specific kind of website, to a specific kind of competitor, while still running an enormous share of the serious web. If you own a WordPress site, the chart matters less than which side of that split your site sits on — and that is something you can actually check.
The number everyone is quoting
The figure at the centre of this debate comes from W3Techs, the long-running technology survey. On 30 October 2026, W3Techs put WordPress at 39.5% of all websites, and 58.2% of websites that use a content management system it can detect. Its peak in this series was 43.6% of all websites at the start of 2025. On 1 January 2026 it was 43.0%. So the fall from the start of this year to late October is 3.5 percentage points, and from the peak it is just over four points.
That is a genuine, sustained slide, and the first of its kind in WordPress's history. A platform that spent years at around 43% does not drift down to 39.5% by accident.
But context changes the meaning completely. WordPress at 39.5% is still, by a distance, the biggest platform on the web. The nearest rival in the same W3Techs table, Shopify, sits at 5.3% of all websites; Wix is at 4.2%. WordPress is not being overtaken. It is shrinking while staying seven times larger than its closest rival — a dominant platform losing share at the edges, not a death spiral.
There is also a stranger number in the same table. W3Techs has a category called "None", meaning sites where it cannot detect a monitored CMS. That category grew from 28.6% to 32.1% over the same months WordPress fell. The biggest gainer is not a product at all. It is the absence of a detectable CMS.
The caveat I refuse to leave out
Before anyone builds a business decision on this chart, there is a measurement caveat that cuts both ways. W3Techs introduced new CMS survey categories on 30 March 2026, and one analysis argued that roughly two-thirds of the twelve-month decline falls after that change — so some of the movement may be better detection rather than sites actually leaving WordPress.
But that caveat does not rescue the "nothing is happening" argument: WordPress's share of the CMS market itself — counting only sites where a CMS is detected — has also fallen, from around 60% to about 58.4% over twelve months, and it has declined every year since its 2022 peak. That second percentage already excludes the "None" bucket, so detection changes cannot explain it away. The honest summary: part of the headline drop may be measurement, but a real decline remains underneath it. The chart is noisy, not fake.
So is AI killing WordPress?
The short answer is: AI is killing a certain kind of WordPress site, and that kind was always the most fragile part of the platform's growth.
Think about what a simple brochure website used to require: hosting, a WordPress install, a theme, an evening wrestling with a page builder, and five pages nobody would ever update. That was a lot of friction for a site that did almost nothing. Today an AI builder can produce a finished-looking version in fifteen minutes, in a browser, with no hosting dashboard and no plugin updates. For that job, WordPress was always overkill. AI just made the overkill obvious.
The numbers support that reading. Lovable, the AI builder most often named here, runs just 0.3% of all websites in W3Techs data — slightly ahead of Astro and GitHub Pages at 0.2% each, but nowhere near WordPress. AI builders are not replacing WordPress site for site. They are capturing new, simple sites that would once have defaulted to WordPress, while AI, static-generator, framework, and headless output shows up as "None" because no CMS fingerprint is left on the page.
A longtime WordPress user quoted in a November WordPress newsletter put it the way I would: simple sites that never needed a content manager are leaving for AI-built and static options, while sites that do real work — memberships, WooCommerce stores, FluentCRM setups — have little reason to move. The same summary cited Barn2, a well-known plugin company, with new sales down 48% over two years. I treat one company's sales as a signal, not proof, but it fits the pattern: the easy, one-off end of the market is getting thinner.
What WordPress is doing about it
It would be a mistake to read this story as WordPress standing still. WordPress 7.2 is in beta, with a final release scheduled for 8 December 2026, continuing work on the Abilities API, a more flexible REST API, and the admin experience. More importantly, the WordPress AI team has approval in principle to bring MCP — the Model Context Protocol, the open standard that lets AI apps talk to other software — into core, with version 7.3 floated as the target. I covered that decision last month, and the key point still stands: WordPress is trying to become the system AI agents work through, with permissions, rather than the system they work around.
Elementor is moving the same way, with its own MCP beta opening in September 2026. I would still test any agent feature on staging first — an agent editing a global template can break every page at once — but the direction is clear. The builders that survive will be the ones an agent can use safely, not the ones that lock agents out.
None of this guarantees a recovery — a roadmap is not a result — but it does mean the "WordPress is ignoring AI" story is out of date.
The part of the story nobody wants to hear
There is an uncomfortable part of this debate that has nothing to do with AI. Performance and maintenance still decide whether a WordPress site keeps its visitors. One analysis citing HTTP Archive Core Web Vitals data from July 2026 put WordPress sites at 49% passing on mobile, against 53% for the web overall, with good largest contentful paint at 56% against 65%. Those are not catastrophic numbers, but they describe a platform where a heavy theme, a dozen plugins, and unoptimized images are still the default build. An AI builder shipping a fast, simple page will beat a slow WordPress site in search and in user patience.
Security is the same story. WordPress 7.1.4 shipped on 28 October 2026 fixing flaws including an issue in the upgrade process itself, and a 2026 survey found 75.5% of ecommerce respondents had experienced at least one security incident. That does not mean WordPress is unsafe — it means a popular, powerful system punishes neglect, and neglect is common. A site that loads slowly, never updates, and has not published in months will lose to anything, AI or not.
What I would do this week if WordPress was my site
I am not going to tell you to migrate, and I am not going to tell you to stay. Here is what I would check, in order, before making either decision.
First, what does my site actually do? If it publishes regularly, sells products, takes bookings, or runs a membership, WordPress is doing a job simple AI builders do not do well. If it is five static pages unchanged in a year, I would price a rebuild against what the site earns, not what it cost to build.
Second, fix speed before debating platforms: measure mobile Core Web Vitals, remove unused plugins, and compress images. A fast WordPress site removes the easiest reason a visitor prefers a competitor.
Third, update and lock down the basics: run the 7.1.4 security release if you have not already, keep a tested backup you know how to restore, use a dedicated low-privilege account for anything automated, and never hand an AI agent administrator access on a live site. That is the cheapest insurance in this whole story.
Fourth, only consider migrating if a specific, named job is genuinely easier elsewhere, counting redirects, SEO history, content migration, retraining, and lost features honestly. "WordPress is dying" is not a migration plan. A spreadsheet with one job, one cost, and one risk is.
Finally, if I built sites for clients, I would move upmarket with the market. The race to build the cheapest five-page site is one AI builders have already won. The work that still pays is what they cannot do alone: content operations, commerce, memberships, performance, and security. That is where WordPress still earns its keep.
Conclusion: not dying, but no longer the default
So, is WordPress dying? No. Is it still the automatic default for every website? Also no, and that second answer is the important one.
WordPress at 39.5% of all websites and 58.2% of the CMS market is still the biggest platform on the web by an enormous margin. What it has lost, in 2026, is the assumption that every new site will be a WordPress site. The simple end now has faster, easier options, and some of those sites will never touch a CMS at all. The measurement is noisy, the decline is real, and neither fact alone tells you what to do.
The sites that will be fine do real work, load fast, stay updated, and treat AI as a tool to test on staging. The sites that will disappear were already neglected — AI just made the exit quicker.
I am staying with WordPress, eyes open. Its job now is to prove, release by release, that a managed, open, permissioned home for content and commerce is worth more than a fifteen-minute generated page — one fast, well-maintained, useful site at a time, starting with yours.
FAQ
Is WordPress really dying?
No. W3Techs put WordPress at 39.5% of all websites on 30 October 2026, down from a 43.6% peak at the start of 2025, but it still holds 58.2% of the CMS market and is roughly seven times larger than Shopify at 5.3%. It is losing share, not disappearing.
Is AI the reason WordPress is losing market share?
Partly, and indirectly. AI builders capture simple new sites that once defaulted to WordPress, and AI-generated output often shows up as "None" because no CMS is detectable. But builders like Lovable, at 0.3% of all websites, are far too small to explain the whole decline alone.
Should I migrate because of the market-share decline?
Not on the chart alone. Migrate only if a specific job is genuinely easier elsewhere, after counting redirects, SEO history, content migration, and lost features. Sites that publish, sell, take bookings, or run memberships still have strong reasons to stay.
What does the "None" category in W3Techs mean?
W3Techs could not detect a monitored CMS. That includes hand-coded HTML, static generators, frameworks like Next.js, headless setups, and AI-generated pages with no CMS fingerprint. It does not mean those sites run on nothing.
What should WordPress site owners do first?
Fix the basics: measure mobile Core Web Vitals, remove unused plugins, compress images, apply the latest security release, keep a tested backup, and test any AI workflow on staging with a low-privilege account — never as administrator on a live site.
Does this change anything for Elementor users?
Not the fundamentals. Elementor remains a major way WordPress sites get built, and its MCP beta points toward agents working through the builder's structures. The shift is toward higher-value work — commerce, memberships, content operations, performance — not the cheapest five-page builds AI tools now do in minutes.
