For two decades, building a website meant building it by hand: pushing pixels in a design tool, then translating every box and margin into code or a page builder. That part of the job is disappearing fast. Not because websites stopped needing designers — but because the work designers are valued for is moving somewhere else.
Key takeaways
- Building websites by hand is fading — AI can now produce layouts, components and code in minutes.
- The designer's role is shifting from making to directing: strategy, taste and judging the output.
- MCP (Model Context Protocol) lets AI assistants work directly inside tools like WordPress, Elementor, Webflow and Framer.
- Taste and strategy are still human jobs — and they're what separates a great website from a generic one.
Is building websites by hand really dead?
Mostly, yes — at least the “by hand” part. Today an AI assistant can set up a page structure, write the CSS, create components and fill them with content faster than any of us can drag a container onto a canvas.
What isn't dead is the thinking behind a website. Someone still has to decide what the site is for, who it speaks to and what “good” looks like. That someone is still the designer — just in a different role.
How the web designer's role is changing
The shift is from builder to director. Instead of producing every element yourself, you brief, review and steer an AI that does most of the production work.
Where designers used to spend their time
- Translating mockups into layouts, pixel by pixel
- Rebuilding the same components on every project
- Fiddling with responsive breakpoints and spacing
- Writing and debugging front-end code
Where designers spend their time now
- Defining the strategy and structure of the site
- Setting the visual direction: type, colour, spacing and tone
- Briefing the AI with clear goals and good references
- Reviewing the output and pushing it from “fine” to “great”
Taste is the new technical skill
When anyone can generate a website in minutes, the hard part isn't producing one — it's knowing whether it's any good. That takes taste: a trained sense of what good looks like, and why.
AI is very good at producing something plausible. It's much worse at telling plausible apart from excellent. If you can't spot the difference yourself, you'll end up shipping whatever it gives you.
When I review AI-generated design, these are the things I check first:
- Hierarchy: is it obvious what matters most on the page?
- Spacing and rhythm: does the layout breathe, or is everything evenly crammed?
- Typography: are there real contrasts in size and weight, or is everything medium?
- Colour and contrast: is it accessible in both light and dark mode?
- Consistency: do buttons, cards and headings follow the same rules everywhere?
- Message: does the copy say something specific, or could it be anyone's website?
What is MCP (Model Context Protocol)?
MCP, short for Model Context Protocol, is an open standard introduced by Anthropic in late 2024 that lets AI assistants connect to the tools you already use. Think of it as a universal adapter: instead of copying and pasting between a chat window and your website builder, the AI can read your project and make changes in it directly — with your permission.
For web design, that means an AI assistant can:
- Read your existing pages, design tokens and components
- Create new pages and sections using your own building blocks
- Update styles, colours and typography across the whole site
- Check its own work and fix what doesn't look right
Modern web design tools with MCP support
Most of the big website builders now offer a way to connect AI assistants through MCP, either officially or through plugins. Here's how it looks in the tools I see most often.
WordPress
The most widely used CMS on the web can be connected to AI assistants through MCP, so content, pages, menus and settings can be managed in a conversation instead of clicking through dashboards.
Elementor
With MCP, an AI can build layouts using Elementor's own widgets, global classes and variables — so the result stays fully editable in the visual editor, just like anything you built yourself.
Webflow
Webflow offers an MCP server that connects AI assistants to your sites, so they can work with pages, styles and CMS content.
Framer
Framer projects can be connected to AI assistants through MCP plugins, letting AI read and update your designs and content.
The common thread: the AI doesn't just suggest changes — it builds inside the same tool you'd use yourself. That makes your judgement more important, not less.
The work that's still on the designer's shoulders
Strategy
AI can't sit down with a client and figure out what the business actually needs. Who is the site for? What should a visitor do next? What's the one thing they must understand in five seconds? Those answers shape everything else, and they come from conversations, research and experience.
Training your design eye
Taste isn't magic — it's trained. The best way I know is to look closely at what other designers ship, on galleries like Awwwards, Land-book or Godly and in the products you use every day, and to keep asking why something works.
Borrowing style without copying
Getting inspired is part of the job. The skill is taking the principles behind a style and applying them to your own site, instead of cloning someone else's layout:
- Pick two or three references, not twenty
- Note the principles: type scale, spacing, colour ratios, how images are used
- Turn them into rules: design tokens, components and a short style guide
- Apply them to your own content, so the result fits your brand rather than theirs
What about plain code?
Website builders aren't the only option. AI will write HTML, CSS, React or a full Next.js site without hesitation — and often the code is perfectly fine. But good code isn't the same as the website you actually wanted. You still have to guide it.
What helps most when building a site with AI in code:
- Describe the audience and the goal before asking for any design
- Share references and explain what you like about them
- Define your design tokens — colours, fonts, spacing — up front
- Work section by section instead of asking for a whole site at once
- Review every result against your checklist, then iterate
Can clients build their own website with AI?
In theory, yes. In practice, I haven't yet seen a client build their own website with AI in the wild. When it does happen, I'd expect the result to fall short of what an experienced designer delivers — on both taste and strategy.
The gaps I'd expect to see:
- Generic layouts that look like every other AI-generated site
- No clear message or next step for the visitor
- Everything at the same visual weight, so nothing stands out
- Inconsistent spacing, type and components from page to page
- Accessibility issues nobody noticed
AI makes it possible to build a website without a designer. It doesn't make it possible to build a good one without design thinking.
How web designers can stay ahead
- Get fluent with AI and MCP in the builder you already use
- Invest in strategy: ask better questions before you design
- Train your eye every week by studying great work
- Build reusable systems — tokens, classes and components — that AI can follow
- Treat AI output as a first draft, never the final answer
Frequently asked questions
Is web design dead because of AI?
No. Building websites by hand is fading, but web design — deciding what a site should say, how it should feel and how it should work — matters more than ever. AI handles more of the production; designers handle the direction.
Will AI replace web designers?
AI is replacing parts of the job, especially repetitive production work. Designers who bring strategy and taste, and who know how to direct AI, are becoming more valuable, not less.
What is MCP in web design?
MCP (Model Context Protocol) is an open standard that connects AI assistants to the tools you use. In web design, it lets AI work inside your website builder directly — creating pages, updating styles and managing content.
Which website builders support MCP?
WordPress, Elementor, Webflow and Framer can all be connected to AI assistants through MCP, using official servers or plugins. Support is growing quickly, so check each tool's documentation for the latest options.
Can I build my own website with AI?
You can get a working website, especially for simple needs. For a site that stands out and turns visitors into customers, it still pays to work with someone who has strong design judgement and a clear strategy.
What skills do web designers need in the age of AI?
Strategy, taste, systems thinking, clear communication with AI tools and the ability to critically review output — knowing what good looks like, and how to get there.
The bottom line
Building websites by hand is on its way out. Designing them isn't. The designers who thrive will be the ones who know what good looks like, can explain it clearly, and can steer AI until it gets there.
Want a website with that kind of thinking behind it? See how I design and build websites or book a call.
