Will AI replace web designers?

You’ve had a quote for a new website. Then a mate (or some bloke on LinkedIn with a robot emoji in his name) told you AI will do the whole thing for nothing, or 20 quid for a Claude subscription, so why on earth would you pay anybody now?

I get told a version of this every week now, usually on my Instagram. My simple answer to ‘will AI replace web designers?’ is…no. And the reason I think no has nothing to do with AI being rubbish. It’s actually quite good.

I know, I know. This is a web designer telling you web designers are safe. So I’ll show my workings. I had a first go at this in Is AI stealing my job?, which has more shouting in it than this does.

What can AI already do on a website?

A lot, actually. I use it every day: ChatGPT, Claude, inside Notion, inside Missive, Basecamp. It’ll write code, help you draft copy, mock up layouts and turn sketches on the back of an envelope into a working page in 10 minutes. I find myself using it a lot to help do much more detailed research on website projects before I even start designing.

The bits it’s properly good at:

• First drafts of copy or plans for what needs writing
• The boilerplate code nobody has ever enjoyed typing
• Resizing a load of images in one go
• Writing the alt text nobody was ever going to write anyway
• Doing deep research on a website before we start designing anything
• Claude is getting better at designing websites and helping you with that, but it’s nowhere near perfect. And don’t even get me started with ChatGPT design.

If all you need is one page that says who you are and has a phone number on it, AI can do that today, in less time than it takes to make a cup of tea. But then, if that’s all you need you might just be better with a Facebook page anyway.

I went through the longer version in Can AI build a website for your business?, and there’s a daft sketch about it too, Can you just use AI?

So AI can build a website. That bit’s settled.

So what does a web designer actually do, if AI does the building?

It might surprise you, but typing the code was never really the bit you were paying for. Like solicitors, most people already used boilerplate code that they tweaked and frameworks that gave them the majority of the code in the first place.

Most of a project goes on, and I hate to say these words: thinking and strategy. That means deciding what the website is for and who it’s for, what it has to say in the first five seconds, and what happens after somebody presses send (does the enquiry land in an inbox somebody checks, or in Dave’s old Hotmail that he left in 2019?).

To get a good website, it’s all in the planning. It isn’t in how well it was coded.

AI is a very fast puzzle-solver with no drawings or workings out. Point at a wall and it’ll have a staircase up it by dinner, but it won’t once ask where the staircase is meant to go.

You end up with a lovely staircase to nowhere and a joiner who’s very pleased with himself.

If indeed, AI could build staircases. But you get my point.

AI builds what you ask for. Most businesses don’t know what to ask for (no shame in that, that’s why they come to us).

Nobody’s ever typed “tell me my homepage is all about our journey and nobody cares” into a chatbot, and that’s the sentence most homepages need to hear.

▶Play the video: AI has changed websites forever, but not in the way you think.

The bits AI can’t do for you (yet)

Glasscraft make glass. Loads of it, cut to whatever sizes customers need, with 150+ people in the building. A customer wants a price for their sizes, so the website needed a Quick Quote tool, plus live order tracking so people could see where their glass had got to without ringing up to ask.

Before any of that got built, multiple people had to be available to answer the phone every day and deal with quoting jobs that could have been handled through the website.

But AI could have never worked out all by itself how to build that quoting tool. It required us to go over and see how their team worked, what the real problem was, and how to build something to replace that manual process.

Cannon Hall Farm’s off-the-shelf ticketing couldn’t keep up, so we built them their own, along with a member portal. It’s sold more than £10 million in tickets, and over 200,000 visitors have a digital passport.

You can read how it went on the page about Cannon Hall Farm’s ticketing system. Nobody prompted that into existence over a weekend.

Then there’s the awkward conversation, where I tell a client their favourite page is the problem. AI won’t have that conversation with you.

It agrees with whoever’s typing, so if you ask it whether a 2,000-word history of the company since the year dot makes a good homepage, it’ll tell you it’s a compelling narrative and offer to make it longer.

“You’re absolutely right!”

The conversations, the planning, the organisation of thousands of content documents, the photography, videography…that’s most of what goes into properly made websites, and a chatbot’s no help with any of it.

Will AI replace the cheap end of web design?

Yes. The 500-quid template site, the “we’ll pop your logo on a theme” job, is being eaten alive by AI builders, and as it should be. You’ll get a lot more flexibility and customisation building it with AI than you ever would with Wix.

I’ve had the 500 quid conversation for years. Somebody rings up, loves everything we’ve shown them, then asks if we can do it for 500 quid, because their cousin’s mate’s lad did one for that (I made a video about it, Can you do it for 500 quid?, and it’s barely exaggerated).

Anyway.

If a cheap website was always going to cost you twice, an AI one costs you twice, twice as fast. Buy cheap, buy twice does the sums on that.

The designers who should worry are the ones whose whole job was the typing. If you were selling a theme with somebody’s logo on it, a machine does that now, and it doesn’t need to pay you to do it.

Should I wait for AI to get better before I get a new website?

Not if your current website is costing you enquiries now. Waiting has a price, you just don’t get an invoice for it.

Wolf Laundry were spending £3,000 a month on Google Ads and getting zero leads back, because the website took ten seconds to load. We rebuilt it and got the load time down to about one second. It brought in 30–60 leads a month within weeks of launch, and it now brings in 150–200.

Do the sums for your own business. Every month you spend waiting for a cheaper tool is another month of the old website doing exactly what it’s doing now, which is nothing.

For Wolf Laundry that was £3,000 of ads pointed at a page people gave up on, and a year of waiting at that rate is £36,000 with nothing to show for it.

Not every website needs replacing. If you’re not sure about yours, do you need a new website? runs through the signs. Some don’t, and that’s fine.

And if the worry is that AI will make websites pointless anyway, it hasn’t killed website traffic either. I went through the numbers in Is AI killing website traffic?

I might be wrong about all of this

The tools get better every few months, sometimes every few days right now. I’ve no idea what 2030 looks like, and anybody who says they do is selling a course or lying to you. Sometimes both.

If I’m wrong, the parts of the job that are about thinking and talking to people get bigger, and the machine does more of the typing. I’d cope. I never liked the coding much anyway.

I’ve also been predicting the death of things since 2010 and been wrong most of the time. But I was right about Google Wave. Who was ever going to use that anyway?

If it does take my job, at least it’ll have to sit through the can you make my logo bigger conversation.