On August 6, 1991, a researcher at CERN named Tim Berners-Lee put a single page online. It was gray, it was ugly, and it explained — in plain hypertext — what the World Wide Web was. There was no design, because there were no designers. There was no platform, because the page was the platform. If you wanted a website in 1991, you learned HTML, you found a server, and you typed. Every website on Earth was custom-coded, because there was no other kind.
What happened over the next thirty-five years is usually told as a story of progress: the web got easier, then it got templated, then it got democratized, and now it's getting automated. That's the sanitized version. The truer version is a story about a single economic constraint — the cost of skilled human labor — and the increasingly elaborate compromises the entire industry built to route around it. WordPress was a compromise. Squarespace was a compromise. Wix, Shopify, the page builder, the theme marketplace: all of them were, at bottom, machines for letting businesses avoid paying developers.
Those compromises made sense for twenty years. They are about to stop making sense. To understand why, you have to walk the whole timeline — because the thing AI is disrupting isn't web design. It's the reason web design platforms were ever invented in the first place.
I. The Hand-Built Web (1991–1999)
For the first few years, the web was a hobbyist's medium. In 1993, CERN placed the underlying technology in the public domain and the Mosaic browser gave it a graphical face; by 1994 Netscape Navigator had turned browsing into a mainstream activity. The number of websites went from a few hundred in 1993 to tens of thousands by 1995 to over a million by 1997. Amazon and eBay both launched in 1995 — the same year JavaScript was invented — and suddenly the web wasn't a curiosity. It was a storefront.
But here's what's easy to forget: through this entire explosion, building a website remained a fundamentally artisanal act. Businesses that wanted an online presence had exactly two options. Hire someone who could write HTML by hand, or become that someone. Tools like Microsoft FrontPage (1996) and Macromedia Dreamweaver (1997) softened the learning curve with WYSIWYG editors, and GeoCities gave hobbyists free neighborhoods to homestead, but a serious commercial website was a bespoke construction project. Agencies billed like architects. A corporate site in 1998 could run five or six figures, take months, and break the moment someone in marketing wanted to change a headline — because changing a headline meant calling the developer.
This was the original sin of the early web, and it shaped everything that came after: the people who owned websites could not touch them. Content lived inside code, and code lived inside a specialist's head. The dot-com boom poured money into this arrangement; the dot-com bust, in 2000 and 2001, exposed how absurd it was. When budgets evaporated, thousands of businesses were left holding expensive, hand-built sites they had no ability to maintain.
The market had learned its first lesson: custom code was powerful, but custom code was a liability if you couldn't afford the person who wrote it.
II. The Great Decoupling: WordPress and the CMS Era (2003–2010)
The answer arrived from an unglamorous direction: blogging. Movable Type shipped in 2001, and in May 2003, a nineteen-year-old named Matt Mullenweg forked an abandoned blogging tool called b2/cafelog and released WordPress. It was free, it was open source, and it did one revolutionary thing — it separated content from code. For the first time, a business owner could log into an admin panel and change the words on their own website without knowing what an HTML tag was.
It's hard to overstate what this unlocked. The theme meant design could be bought instead of built. The plugin meant functionality could be installed instead of programmed. A whole economy sprouted around WordPress — theme marketplaces, plugin developers, hosting companies, and an entire tier of "WordPress developer" that was really a new kind of professional: not quite an engineer, not quite a designer, but a skilled assembler of other people's components. The cost of a functional business website dropped from tens of thousands of dollars to a few thousand, and the maintenance hostage crisis of the 1990s largely ended.
The results speak for themselves. WordPress grew from a blogging tool into the default infrastructure of the web, eventually powering more than forty percent of all websites — a market share no software platform in any category has ever matched. Governments run on it. Newsrooms run on it. Fortune 500 marketing sites run on it. For a stretch of about a decade, "we need a website" and "we need a WordPress site" were functionally the same sentence.
But WordPress was a half-measure, and everyone quietly knew it. It removed the developer from content updates, not from the site itself. You still needed someone to set it up, secure it, update the plugins, untangle the theme when it collided with another plugin, and rescue you when the whole thing got hacked — which it did, constantly, because being forty percent of the web makes you a hundred percent of the target. WordPress democratized owning a website. It never really democratized building one.
III. The Template Age: Squarespace, Wix, and the Death of the Blank Page (2004–2020)
The next compromise went further. Anthony Casalena started Squarespace in a University of Maryland dorm room around 2003–2004. Wix launched out of Tel Aviv in 2006. And that same year, a German snowboarder in Ottawa named Tobias Lütke, frustrated that no existing software could run his snowboard shop's online store, built his own — and then realized the software was the business. He called it Shopify.
These platforms made a bet WordPress never quite made: that most businesses don't want flexibility. They want a result. Squarespace and Wix collapsed hosting, design, security, and editing into a single subscription. You didn't install anything. You didn't update anything. You picked a template — invariably beautiful, invariably minimalist, invariably interchangeable — dragged your logo into place, and you were live by dinner. Shopify did the same for commerce, and did it so well that it became the invisible plumbing behind millions of stores, from garage side-hustles to billion-dollar direct-to-consumer brands.
Two outside forces supercharged the template age. The first was the iPhone. After 2007, and especially after Ethan Marcotte coined "responsive web design" in 2010, every website suddenly needed to work on a four-inch screen. When Google's 2015 "Mobilegeddon" update began rewarding mobile-friendly sites in search rankings, responsiveness stopped being optional — and the platforms handled it automatically, which made hand-built sites look even more expensive by comparison. Why pay a developer to make your site responsive when Wix did it for free?
The second force was cultural: the website became table stakes. By the mid-2010s, a business without a website read as a business that might not exist. The pandemic finished the job. In 2020, restaurants, gyms, dentists, and plumbers who had survived for decades on word of mouth suddenly needed online ordering, online booking, online everything — and they needed it in a week, not a quarter. The template platforms absorbed that demand at a scale no agency model could have touched. It was their finest hour.
It was also the moment their limitations calcified into the landscape. Walk down any commercial strip in America — digitally speaking — and you'll see the result: millions of websites that are professionally adequate and completely indistinguishable. Same hero image, same three-column feature grid, same testimonial slider. The template age solved cost and speed, but it did so by industrializing sameness. And it introduced a subtler tax that most owners never itemized: the platform owned the ceiling. Your site could do whatever Squarespace decided sites should do. Your store's checkout worked however Shopify said checkouts work. Custom functionality — the thing that had made the hand-built web powerful — was either impossible, or available only by climbing back into code through a trapdoor the platform grudgingly left open.
For most businesses, most of the time, that trade was worth it. The alternative was five figures and three months with an agency. Against that price, sameness was a bargain.
IV. The Quiet Counter-Current (2013–2022)
While the template platforms conquered Main Street, something else was happening among developers. The professional web was splitting away from the platform web and getting more custom, not less. React (2013) and its descendants rebuilt how serious applications were engineered. The JAMstack movement pushed toward fast, pre-rendered, custom-coded sites served from the edge. Headless CMS architecture re-separated content from presentation, but this time with the presentation layer written by hand again. Webflow, founded in 2013, split the difference — a visual builder that generated real, exportable code, aimed at designers who wanted custom output without hand-writing it.
The signal underneath all of this was easy to miss but important: whenever the cost of custom development came down — through better frameworks, better tooling, better abstractions — demand for custom work went up. Nobody actually preferred templates. Businesses tolerated templates because bespoke was priced out of reach. The preference for custom had never gone anywhere. It was just waiting for the price to break.
V. The Break (2022–Present)
The price broke in about eighteen months.
GitHub Copilot arrived in 2021 as a curiosity — autocomplete for programmers. Then ChatGPT launched in November 2022, and the general public discovered that a machine could write working code from a plain-English sentence. What followed was the fastest capability ramp in the history of software tooling: models like Claude and GPT-4 that could architect entire applications, AI-native editors like Cursor, generative UI tools like v0, and agentic systems like Claude Code that don't suggest code so much as ship it — scaffolding a project, writing the pages, wiring the forms, fixing their own errors, and deploying the result.
Consider what this does to the economics that created every platform on the timeline. The entire template industry exists because custom development cost $10,000 to $100,000 and took months. That was the wall. Squarespace's real product was never templates; it was not needing a developer. WordPress's real product was never a CMS; it was not needing a developer for updates. When an AI system can produce a genuinely bespoke, professionally engineered website — custom design, custom functionality, clean code you actually own — in days instead of months and at a fraction of the old cost, the wall those platforms were built against simply isn't there anymore.
This is the strange, almost poetic shape of the timeline: the web is returning to where it started. Custom code, hand-fitted to each business — except the hand is now augmented by machines, and the economics that made hand-fitting a luxury good have inverted. What was artisanal in 1996 and aristocratic in 2016 is becoming accessible in 2026.
VI. Why This Time the Pendulum Sticks
A fair skeptic would point out that the web has seen "the return of custom" before, and the platforms won anyway. Three things make this swing different.
First, the maintenance problem — the original sin that killed the hand-built web in 2001 — is solved by the same technology that revived it. A custom site in 1999 was a hostage situation because only its author could modify it. A custom site in 2026 can be modified by anyone who can describe the change in a sentence, because AI reads and writes the code. The historical liability of ownership is gone. What remains is only the asset.
Second, the audience for websites has changed in a way templates are structurally unprepared for. Increasingly, the most important visitor to your website is not a person. It's an AI. Google's AI Overviews, ChatGPT's browsing, Perplexity, and the answer engines behind every voice assistant now read the web to decide which businesses to recommend. And they don't judge a site by its hero image. They read its structure: its schema markup, its structured data, its machine-readable declarations of who this business is, what it does, where it operates, and why it can be trusted. A cookie-cutter template gives an AI crawler the same thin gruel it gives every other template. A custom-engineered site can be built, from the foundation up, to be legible to machines — and in a search landscape where the machine's summary increasingly is the search result, that legibility is the new SEO. The template age optimized websites for human eyeballs at exactly the moment eyeballs stopped being the only audience that matters.
Third — and this is the part the platform companies understand better than anyone, which is why they are all racing to bolt AI generators onto their products — differentiation compounds. When every competitor can have an adequate website, adequate is worth nothing. The template age created a sea of digital sameness precisely when standing out became the entire game. The businesses that win the next decade of the web will be the ones whose sites can do things their competitors' sites can't: custom tools, custom integrations, custom experiences, custom data. None of that lives inside a template. All of it lives in code.
VII. The Whole Arc, Seen at Once
Step back far enough and the timeline stops looking like a sequence of technologies and starts looking like a single negotiation, conducted over thirty-five years, between what businesses wanted and what they could afford.
In the 1990s, they got custom code because there was nothing else — and it nearly bankrupted them to maintain it. In the 2000s, WordPress let them keep the website and fire the webmaster. In the 2010s, Squarespace, Wix, and Shopify let them skip the developer entirely, at the price of looking like everyone else and living inside someone else's ceiling. Each era traded away a little more ownership and a little more distinctiveness in exchange for a little more affordability. Each trade was rational. Each one was also, quietly, a loss.
AI is the first technology in the history of the web that doesn't ask for that trade. It offers the craftsmanship of 1998, the ease of 2015, and the machine-legibility that 2026 demands — simultaneously. That has never been on the table before. It's why the smartest builders right now aren't asking which platform to choose. They're realizing the platform question itself belonged to an era that just ended.
The web began as a page a man typed by hand in a physics lab, and for three and a half decades we built ever-more-elaborate machinery to avoid typing. Now the machines type. The circle closes not where anyone predicted — not in no-code, not in drag-and-drop, not in yet another template marketplace — but back at the beginning, in custom code, where the web was always at its best. We just finally stopped having to choose between owning something great and being able to afford it.