"Will AI replace web designers?" is the question we get asked most often now, usually from someone halfway into a conversation about their site, usually phrased as though they have already decided the answer and are just checking whether we have noticed.
Our honest answer is that it will take part of the work, eventually, the way a calculator took long division. But we have heard this exact question before. Twice. Both times the answer turned out to be more interesting than a yes.
We have run this experiment before
Before AI, the same prediction arrived with the same certainty, just wearing different clothes.
First it was templates. Buy a theme for sixty dollars and you were done, no designer needed. For a while that was almost true. What actually happened is that the businesses who bought themes got websites that looked like everyone else's and ranked like everyone else's, which is to say not at all, because a template cannot know that your customers search for a phrase your industry does not use. The templates did not disappear. They became one input into a decision instead of the whole decision.
Then it was outsourcing. Hire a team overseas for a fraction of the price and skip the local agency. That one had more staying power because there is real skill in a lot of those teams and we have worked alongside plenty of them. What it ran into was not ability. It was distance. Not geography, the other kind: nobody in the room who had ever watched a customer use the site, nobody who knew that the owner's brother is the operations manager and that everything routes through him, nobody to call on a Tuesday when the contact form quietly stopped sending leads two weeks ago and nobody noticed.
Both times, the tool was real, the savings were real, and the thing people actually needed turned out to be attached to the tool rather than inside it.
What AI is genuinely good at
We are not going to pretend the current wave is the same as the last two, because it is not.
AI is fast at the parts that are mechanical. It writes the first draft of copy that would have taken a junior writer a day. It produces a working site structure in an afternoon. It generates clean markup, it names things consistently, it writes the boring pages: the privacy policy, the terms, the FAQ that nobody on your team wants to spend a Friday on. It can take a photograph you already own and resize it for four breakpoints.
If your need is a page that exists, that looks presentable, and that you can put up this week, AI will do that. We would not tell you otherwise, and we use it ourselves for exactly those jobs. Fighting that would make us the wrong people to ask.
What it cannot do
Here is where we part company with the enthusiasm, and it is not about quality. The current models produce genuinely good output most of the time. The problem is the rest of the time, and what happens when nobody in the room can tell the difference.
AI does not know when it is wrong. It is confident at the same volume whether it is right or inventing something. We have watched it produce a page of clean, plausible, well formatted copy built around a service the business does not offer, because the business mentioned something adjacent once on a podcast. We have watched it write a phone number that no one has ever answered. We have watched it decide that the right headline for a medical practice was funny.
Catching that requires someone who knows what right looks like for this business, this market, this owner. That is not a technical skill. It is the accumulated version of every mistake you have already paid for.
There is a second thing, and it is the one we would put first if we were being blunt. Taste.
Taste is not decoration. It is knowing which of the four versions in front of you is the one, and being able to say why, and being right often enough that people trust the next call. It is knowing that the offer should go above the fold here and below it there, that this competitor's site looks expensive because of one specific choice you can name, that the owner's instinct to put his photograph on the homepage is right for his customers and wrong for his industry.
You cannot prompt for that, because the prompt is the answer. If you already knew what to ask for, you would not need to ask.
The SEO part nobody mentions
Web design is a small part of what actually determines whether a site works. Most of it is search visibility, which is a longer and stranger discipline than the layout.
An AI will happily write you a page titled "Services" and another called "About Us" and consider the job done. The reason that matters is not aesthetic. It is that nobody searches those phrases. They search "emergency plumber near me at 2am" and "why is my electric bill higher than my neighbor's". Finding the difference between what a business calls its work and what its customers type into a phone is a research job, and it is the one where AI most often produces confident nonsense, because it can generate a keyword list instantly and has no way to know which of those keywords a real person has ever typed.
We have spent the last few years watching AI reshape search itself, which makes this harder rather than easier. If you want the long version of that, it is here on the site. The short version: the businesses getting found are the ones whose content answers a specific question in a specific person's words, and producing that is editorial judgment, not generation speed.
If you are a business owner deciding what to do this month
Three things we would actually do, in this order.
Write down what your site is for, in one sentence, before you let anything generate anything. Not "to have a web presence". Something like "to get a phone call from a homeowner in Decatur whose AC died on a Sunday". Every decision after that gets easier and most of them get obvious, because now there is a test to apply.
Use AI for the parts that are already decided. Once you know what the page must say, let it write the draft, let it write the boring pages, let it resize the images. That is speed with no downside.
Get one human pass on anything a customer will read, from someone who will say no. This is the step people skip because the output looked fine, and it is the step that catches the phone number nobody answers and the service you do not offer. Fine is not the bar. Correct is the bar.
Where this leaves us
We are not going to argue that AI is a fad, because it is not, and we would rather be useful than comfortable. We are going to argue that the value has moved, and that it moved to the part people keep describing as the soft stuff.
A machine can build a website now. What it cannot do is look at what it built and tell you that this one is going to make you money and that one is going to quietly embarrass you in a year, because that judgement comes from having been wrong in public and paying for it.
We have been building, hosting and looking after websites. Before that, our founder was writing software on machines most people have never heard of, running bulletin board systems in the late eighties, and working with companies you would recognise long before any of this was a category. That is not a sales line, it is the reason we can look at a generated page and tell you, with specifics, what is wrong with it.
Experience still matters. It just moved to the part of the work that was always the hard part.
If you would like a second opinion on a site you already have, or a straight answer on whether AI can do the job you have in mind, talk to us. We will tell you what we think, including when the answer is that you do not need us for this one.