Everyone’s got an opinion on AI and web design right now. Most of them are based on either hype or fear.
Mine is based on actually using it — properly, day in and day out, across client projects, experimental builds, and personal work.
I’ve been building websites for fifteen years.
For the last several months, Claude Code has been part of my daily workflow. I’ve used it for scaffolding themes, building custom components, cloning layouts, generating WordPress templates, and producing full scroll-driven 3D sites from a single prompt.
I’ve pushed it hard. I’ve broken it. I’ve built on top of it. And I’ve watched other people use it with very different results.
So here’s the honest version of what’s going on, from someone who’s not selling an AI tool and not shouting that the sky is falling.
Looking for something in particular? Jump to one of the following sections:
Claude Code Is Genuinely Brilliant — If You Know What You're Doing
Let me say the positive bit first, because it matters. Claude Code is one of the most useful things to land in my workflow in years.
It can spin up a theme file structure in seconds. It can write CSS that matches a design brief closely enough to refine rather than rebuild.
It can take a messy client requirement and give you a starting framework to work from.
When I need to prototype something — a landing page, a product page, an interactive component — I can go from idea to working draft in minutes rather than hours.
Used well, it’s a serious force multiplier.
I can take on more work, ship faster, and spend the time I save on the things AI can’t do — strategy, client relationships, creative direction, proper problem-solving.
That’s the caveat though. Used well.
The Problem Isn't AI. It's AI In The Wrong Hands.
Here’s what I keep seeing.
Business owners, marketers, and people with no development background are using AI tools to generate entire websites. Then pushing them live. Wondering why things aren’t working.
And on the surface, these sites look fine. That’s the trap. They render in a browser. The images load. The buttons click. Job done, right?
No. Not even close.
A website isn’t a pretty page. It’s a piece of business infrastructure.
It needs to rank on Google. Convert visitors into enquiries. Handle real traffic. Integrate with your CRM. Comply with accessibility law. Stay secure.
And be maintainable when something breaks or needs updating six months down the line.
AI, in my experience, gets about 40% of the way there. The visible 40%.
The other 60% is where everything quietly falls apart.
What I Keep Finding Under The Bonnet
I’ve now looked at enough AI-generated sites to see the same problems show up again and again. Here’s the list:
Why this Is good news for web designers
Here’s where I disagree with most of the doom-and-gloom posts you’ll see on LinkedIn about AI killing web design.
It hasn’t. It’s done the opposite.
The volume of rescue work landing on my desk has been relentless.
Clients who generated a site in a weekend, pushed it live, and six weeks later realised the leads weren’t coming in. The rankings had disappeared. Or something important had broken.
This market didn’t exist two years ago. Now it’s one of the busiest parts of what I do.
If you’re already sat with an AI-built site and something feels off, you can grab a free audit — I’ll take a proper look and let you know what’s going on. More on that at the end of this post.
People are building more websites, faster, than at any point in history.
And when you build that much, that quickly, with tools most users don’t fully understand — you create an enormous amount of cleanup work.
Bad architecture. Non-converting pages. SEO disasters. Security holes. Sites that break under real traffic.
Someone has to come in and sort it out. That’s the opportunity.
Not competing with AI — cleaning up after it.
Architecture that doesn't scale
The code looks plausible but the decisions behind it often make no sense.
Nobody — not even the person who prompted it — can explain why the site is structured the way it is.
The moment you try to add something the original prompt didn’t anticipate, the whole thing fights back.
Sites that don't convert
AI defaults to the most statistically common layout it’s seen.
Hero section, three feature cards, testimonials, pricing, CTA. Every. Single. Time.
There’s no thought about what actually makes your audience buy. What objections your sales team hears on calls. What your particular market responds to.
It’s a template with your logo on it.
Weak SEO underneath the surface
AI is great at the checklist — meta tags, heading structure, alt text. Tick, tick, tick.
What it can’t do is the strategic layer. Knowing which keywords actually matter for a plumber in Chelmsford versus a SaaS company in London.
Or why your product page needs to directly answer a specific question your customers keep asking.
That’s not pattern matching. That’s understanding a business.
Security holes nobody's checking
AI will happily wire up a form, an integration, or an authentication flow that technically works but hasn’t been hardened.
I’ve seen exposed API keys, missing input validation, and contact forms that are effectively open doors for spam.
The person who generated the site doesn’t know to check — because they don’t know what they’re looking for.
Accessibility failures
95% of websites don’t meet basic WCAG standards, and AI-built ones are usually worse, not better.
The semantic structure is often wrong. Colour contrast gets ignored. Focus states don’t exist.
And with accessibility lawsuits climbing fast in the UK and EU, this isn’t theoretical — it’s a real risk.
Maintenance nightmares
When something breaks — and something always breaks — you’re now reverse-engineering code written by a model that didn’t know your business, your payment flow, or your integrations.
Good luck.
At least with a bad plugin combo you can toggle things off one at a time. AI-generated architecture doesn’t give you that.
Where I Think The Industry Is Heading
A few predictions, for what they’re worth:
AI will keep getting better. Of course it will.
The gap between “renders in a browser” and “runs a business reliably” will close over time.
But right now that gap is enormous. And it’s not closing as fast as the volume of generated sites is growing.
So the rescue market is going to get bigger before it shrinks, not smaller.
The designers and developers who thrive will be the ones who use AI as a tool rather than fight it.
I use Claude Code every day. It makes me faster, not obsolete.
The ones being replaced are the ones who were only doing what AI can already do — pumping out generic templates with their name stamped on them.
Strategy, brand, conversion, and business context will become the most valuable things a web designer offers.
The mechanical parts of the job will increasingly be automated. The thinking parts can’t be. And those are what clients actually need.
And a whole new category of services is emerging — AI website rescue, automation auditing, AI security reviews, accessibility remediation, and brand differentiation for businesses that accidentally made themselves look like everyone else.
These weren’t jobs two years ago. They’re real jobs now.
If You've Got An AI-Generated Site
Look, I’m not here to tell you that generating a website with AI was a mistake.
For some projects, it’s absolutely the right call — prototypes, test pages, quick internal tools, throwaway campaigns. I use AI for that stuff myself.
What I’d push back on is assuming the output is production-ready because it looks right.
It almost never is.
And the cost of finding that out — when you’re losing leads, getting flagged for accessibility, or watching your rankings tank — is a lot higher than the cost of catching it up front.
If you’ve generated a site with AI and something feels off, head to kubestudio.co.uk/free-website-audit and send it over.
Even if it seems fine but you want a second opinion before you commit to it properly — same thing. Send it in.
I’ll personally go through it and come back to you with what’s working, what isn’t, and what needs to change.
All so your site actually performs the way a proper business website should.
No pitch, no pressure.
Just a real set of eyes on your site from someone who’s spent fifteen years building them — and the last year watching very closely what AI does to them.
