Website Builders Were Made for People. What Happens When AI Is the User?
A website builder is a tool that lets you create a site without writing all the code by hand. Instead of HTML, CSS and JavaScript, you work with ready-made blocks, sections, templates and settings, often just dragging elements around the page.
In essence, someone has already written all that code for you. They left you buttons, fields, sliders and blocks and said: there you go, now build your site.
It’s actually a good idea. At least until you need to do something the builder’s authors didn’t foresee.
Key takeaways
- Website builders turn code into buttons for people. AI doesn’t need the buttons: it can write the code itself.
- Writing code fast and building a site that lasts for years are different things.
- AI helps an expert who can judge its answer, and misleads someone who can’t.
- Without rules, AI solves the same task differently each time, and a project ends up with two systems for one job.
- For AI, a builder’s real value is its rules: how the site is built, what already exists, what is allowed.
Built for people
Over the years I’ve often been asked to fix something on sites made with different builders. And almost every time the same problem came up: before changing a small detail, I first had to understand the logic of that particular builder.
Where is the setting I need? Why is this spacing set here and not there? Why is one block nested inside three more containers? What breaks if I change this?
And almost every time I caught myself thinking the same thing: I’d rather be asked to make sense of an old program written in Delphi.
Because even someone else’s code, in a language you don’t know at all, usually has a clear logic: there’s a condition, a function, data, a result. And most importantly, you can always search code. Find the text, class, function or variable you need, see where it’s used, and gradually work your way to the place you have to change.
In a website builder, it can be the other way round.
You can spend half an hour looking for why a button sits 12 pixels lower than it should, only to find that it depends on the setting of a container nested in another container, and that container’s setting lives on a different page altogether, somewhere in global styles you didn’t even know existed.
The whole construction has always puzzled me a little: developers write an enormous amount of code so that users don’t have to work with code directly. But in return the user gets another complex system, with its own rules, settings, limits and logic that also has to be learned.
Then AI arrived
MCP servers appeared, letting AI do more than answer in a chat: it could work with real tools. Quite quickly the idea reached WordPress and website builders. AI was taught to create blocks, change settings, edit content and work with pages.
In other words, instead of a person clicking the buttons in a builder, the AI started clicking them.
But there’s one important point here. All those panels, drag-and-drop, nested menus, dozens of toggles, the whole logic of the interface, were built around how a person works with a site.
AI doesn’t need any of it. It doesn’t need to drag a block with a mouse. It doesn’t need to open five tabs to change one setting. Even some of the abstractions website builders put on top of code may be unnecessary for AI. Many of them exist precisely so that a person can build a site without working directly with HTML, CSS, JavaScript or PHP.
AI, on the other hand, can write that code itself. So there’s little point making it go through a complex middle layer just because that layer was once needed by a person.
It’s easy to conclude from this that if AI can write code, maybe builders aren’t needed at all. But if you’ve ever seen AI write code with no constraints, you probably understand why it isn’t that simple.
AI really can produce hundreds or thousands of lines of code very quickly. But writing code fast and building a good site that can be maintained, changed and grown for years are completely different things.
How good is AI, really?
Before talking about what a website builder should be in the age of AI, it’s worth talking a little about AI itself, and how well it really helps us at work.
As I write these lines, my girlfriend, who is a doctor, is sending cardiograms to ChatGPT to see what conclusion the AI reaches, and then comparing it with her own. She doesn’t treat its answer as a finished diagnosis. For her it’s more like one more opinion she can set against her own experience.
Sometimes the AI notices a detail that’s easy to miss, or simply makes her look at the result again from a different angle. But the key point is that she understands what she’s looking at, and can judge where the AI is right and where it’s wrong.
I think many of you have run into something similar in your own work. Maybe some of you are lawyers, accountants, designers or specialists in other fields. AI can help a great deal when you understand the subject yourself. You can spot a mistake quickly, ask the right follow-up question, reject a bad solution or steer the AI in the right direction.
But if you don’t know the field at all, the situation changes. AI can give a very convincing answer that looks right at first glance. And without enough experience, you may simply not notice that something in it is wrong.
In my view, this is exactly what matters when we talk about AI that builds websites.
With Fabrement, the user is no longer the source of code, but of judgment. You connect your Figma token and your AI agent, say “do it”, and wake up to a website ready for review: AI can turn Figma into WordPress overnight.
Maksym Verentsov, Head of Fabrement
The same mistakes, in code
In web development, AI makes the same mistakes as in any other profession. It has seen an enormous amount of code, but that doesn’t mean all of that code was good, modern, or right for your situation.
It’s a bit like a law that was once in force but has since been changed or repealed. AI can rely on information that was correct when it appeared but is no longer current. Something similar happens with code: AI has seen countless ways to solve the same problem, modern and outdated, good and bad, simple and needlessly complex. Some of them only worked under specific conditions or on old versions of a technology.
And there’s another problem. Even within one project, AI can make completely different decisions at different moments.
Say yesterday you and the AI built a site and decided to use ACF for the fields and data structure. Today a colleague opens the same project, starts the AI and asks it to finish something. If the AI doesn’t notice that the project already uses ACF, it may well solve the same task its own way: write its own fields, create a new block type, or build yet another system to store the same data.
And the worst part is that both solutions can work. The project simply ends up with two different approaches to one and the same task.
That’s why it isn’t enough for AI to be able to write code. It needs more than the project’s context. It needs the expertise of an experienced developer: knowing which decisions have already been made, what should be reused, where not to create something new, which approaches count as right in this project, and what a change can lead to.
A builder made of rules
And this is where the builder takes on a completely different role.
Its main value for AI is no longer buttons, drag-and-drop or the abstractions that let a person avoid writing code. Its value becomes rules.
A builder can define how a page is structured, how blocks work, where data is stored, how styles are used, which solutions already exist, and what is allowed at all. It can stop the AI from inventing a new structure, a new style system or a new way of solving the same task every time.
And that expertise of an experienced developer is, in my view, exactly what a builder should pass on to AI.
So perhaps the real question is no longer whether website builders will survive in a world of AI. It’s what a builder should be when one of its main users is not a person, but AI.
My answer: a builder for AI is less an interface and more a set of rules. It knows how the site is built, what already exists, and what is allowed, and it makes the AI follow that every time. Buttons were for us. Rules are for the AI.
Andrian Domashovets is the tech lead at 7Fridays and Fabrement, where he builds the MCP server that lets AI agents build and run WordPress sites. Andrian on LinkedIn
Related articles
Why Some Standards Survive and Others Don’t
I lost most of the team standards I tried to keep across three teams. Two held for eight years. What saved them…
The Best CMS Is the One You Don’t Need to Know
Every CMS argument is about which one is best. The real problem is that they are all different, and the knowledge of…
WPVibe vs Fabrement
WPVibe gives an AI agent hands on the WordPress site you already have. Fabrement builds the whole site, from Figma to forms,…