I use AI every single day to build websites. So when people ask me whether AI is going to replace web designers, I'm not going to tell you we're all done for, and I'm also not going to pretend nothing has changed.
Here's my honest answer, and it has two sides. AI changes the job a lot. It's also badly overhyped, for design and for most other things. And the people it really replaces are the ones who were only ever assembling templates, not the ones doing the actual work of design. The job is changing, but it isn't going away. I want to walk through what actually shifts, because I have a very current view of it right now.
What AI is actually good at
I lean on AI constantly, so let me be fair about what it's great at.
It's fast at first drafts. When I'm building, I'll happily let it write the boilerplate, the repetitive scaffolding every project needs and nobody enjoys typing out. I use it to kick off research and to get a rough first pass at content, something on the page I can react to and shape. It's a brilliant starting point, and it saves me real time.
What I don't hand over is the part that decides whether the work is any good. The design direction and the taste stay with me. The final call on whether something is ready to ship stays with me. AI gives me a faster start. I still do the deciding.
Generating a page is not the same as shipping a real site
This is where the hype and the reality split apart, and I'm seeing it up close right now.
I have two projects on my desk at the moment, both apps that were built entirely by AI, start to finish, by someone without the engineering judgment to check the output. The customers are furious. Nothing quite works. The flows are confusing, the interface is baffling at every turn, and the whole thing feels off in the way AI-built work often does.
Then I went in to actually fix them, and what I found was worse. Personal data sitting exposed. Ways into the app that a curious teenager with free AI could find in an afternoon. The kind of holes that turn into a real problem the moment anyone with bad intentions looks.
This isn't only my bad luck. It's a documented pattern. Researchers who tested 1,645 apps built with one popular AI builder found that roughly one in ten were leaking user data through missing permission checks. A journalist wrote about repeatedly stumbling onto AI-built apps that exposed everyone's names, emails, and private messages through a single sloppy line of code, without even going looking for them (XDA). One ID-verification app leaked around seventy thousand images, including thousands of passports and driver's licenses.
The research backs it up. One well-known study found that about 40% of the code GitHub's AI assistant generated across 89 security tasks contained exploitable vulnerabilities (Asleep at the Keyboard?). A Stanford study went further and found that people using an AI assistant wrote less secure code than people working without one, and, worse, felt more confident that their insecure code was safe (Stanford).
The false confidence is what gets people. AI writes fluently, so its output looks finished, and looking finished and being finished are two very different things. I teach, so I watch people run into this constantly. The code runs, the page renders, it all seems fine, and then you look underneath and the actual work hasn't been done.
What AI still can't do
Take away the speed, and you're left with the parts of the job AI can't do for you. Four of them matter most to me.
Taste. Knowing what actually looks right, beyond what's technically correct. AI can produce a clean, reasonable layout in seconds, and it will still miss the feeling a brand needs, because it's working from the average of everything it has seen.
Judgment. Knowing what actually matters for this specific business. Which page matters most, what to cut, where to spend attention. AI treats every request the same. A person weighs it.
Understanding the client. Who they are, how they talk, who they're for, what makes them them. That comes from listening to a real human, not from a prompt.
The engineering to make it real. The work that makes a site actually fast, safe, accessible, and built to convert, which, as the apps on my desk keep reminding me, is exactly the part an AI-only build skips. Accessibility alone is a good example: nearly 96% of the top million websites still fail basic checks that decide whether disabled people can use them at all (WebAIM), and that's the kind of thing you only get right on purpose.
There's also the plain fact that a lot of AI work has a look. You've seen it. That slightly generic, seen-it-before polish. Once you can spot the AI-slop look, you can't unsee it, and neither can your customers.
And here's my own proof of all this. I'm not a designer by trade. I'm a programmer, I used to work in game programming, and I came to websites from the code side. Most of the design side I learned from my husband, who does design and marketing for a living. Even with everything he's taught me, and with AI helping me every day, I still ask for his eye all the time. I can't do the whole thing alone.
If that tells you anything, it's that taste and judgment are real skills that take real people. If AI could do this part for me, I'd have handed it over a long time ago.
The parts that get more valuable
Here's what the scary headlines skip. When AI takes over the fast, repetitive parts, the human parts don't shrink. They become the whole job.
Taste, judgment, understanding people, and building something that actually works were always the hard parts. Now they're the parts that separate a real site from a generated one. The value moves to the decisions, not the typing.
The job numbers point the same way, for what it's worth. I even checked whether they had accounted for AI, since I wondered the same thing, and they had. The US Bureau of Labor Statistics built expected AI impacts into its latest 2024 to 2034 projections, and it still has web development and design growing about 7% over the decade, faster than the average job (BLS). Its read is that AI trims some design and admin roles while raising demand for the people who build and work with the tech. Projections aren't promises, but the folks whose actual job is forecasting this see the same thing I see from my desk. The work is changing shape, not disappearing.
And I'll say this as someone who uses AI all day: I'm more optimistic because of it. It clears the boring parts out of the way, so I can spend my time on the things that actually decide whether a website is good.
If you're hiring someone to build your site
So what does this mean if you're the one paying for a website?
An AI tool, or someone who only knows how to prompt one, can get you something that looks like a website. Quickly, and cheaply. If you're testing an idea and nothing is really at stake yet, that can be enough.
But if the site has to bring in customers, show up in Google, keep people's data safe, and look like you instead of everyone else, you want a person who owns the judgment and uses AI as a tool underneath. Someone who can look at what the AI produced and know what's wrong with it.
I know this because it's a real part of my work now. People come to me to fix sites and apps that were built by AI alone, once the plausible-looking version turns out not to hold up. It works out a lot cheaper to have someone who knows what they're doing from the start.
So, will it replace us?
No. AI is powerful, and I'd feel slower without it. The job was always about knowing what to make, who it's for, and making it actually work. Generating a page was never the hard part. AI changes how I do that work every day, and it still doesn't do it for me.
So I'm not worried. I'm busy, mostly cleaning up after the version of this that skipped the human part. 🐾