AI Built You a Website in Three Minutes. I can tell you if it works or not.

Everyone's Quoting the Wrong Number About AI Websites

You can now describe your business to a prompt box and have a live website before your flat white goes cold. Colours, layout, copy, a contact form, the lot. It will look fine. Genuinely fine. Better than several sites I have been paid actual money to rescue, a sentence I have made my peace with.

The number everyone quotes for this market is 6.3 billion US dollars, and it is worth following that one home, because it lives somewhere embarrassing. It comes from a website builder company's own blog, which credits it to Precedence Research. Precedence's actual published figure is 3.24 billion for 2026. So the number doing the rounds is close to double the research it claims to be built on, and the people circulating it sell AI website builders.

I don't think anyone is lying. I think nobody is checking. Which is the whole problem with "AI built my website" as a phrase. It sounds like information and it isn't. If an industry can't agree on the size of itself within a factor of two, then telling a customer your site was AI-built tells them nothing about whether it works. It is just the only part of the process anyone can see, so it is the part people say out loud. Like telling someone your house was built with a nail gun.

It doesn't help that nobody is measuring the same thing. "AI website builder" was never a product category. It is a feature that got bolted onto everything at once, like USB ports on a fridge. Wix has it. Squarespace has it. Duda, the platform we mostly build on, has it. Draw the boundary tightly and you get three billion. Draw it generously and you get six. Same market, different tape measure.

So here I am. A human who charges actual money to build websites, watching a machine do the visible part of my job in the time it takes to boil the jug. Here is what I am not worried about, and what I am.

Speed was never the bottleneck

Nobody in web design says this out loud, so I will. Making a website look good stopped being hard years ago. A competent person with a decent template and a free Sunday could already produce something respectable in 2019. AI did not crack a problem the industry was stuck on. It sped up the part that was already easy.

The part that was never easy is the part nobody can see. It is knowing which of your five services actually earns its keep, so that one goes at the top and the other four go on a page underneath. It is knowing that the person landing on your site at 9pm on a Sunday is not ready to book, they are ready to compare, and if all you have given them is a Book Now button you have handed them to whoever bothered to publish a price range.

None of that comes out of a prompt. Not because the AI is not clever enough. Because it has never met your customers. It does not know that half your work walks in through one referral partner, or that everyone asks the same question on the phone right before they commit.

It knows what a website for a business like yours usually looks like.

Which is exactly the problem.

Usually is doing an enormous amount of work in that sentence

Ask an AI builder for a site for a mobile mechanic in Silverdale and you get a website for a mobile mechanic. Any mobile mechanic. Anywhere on earth.

Same three services. Same About Us paragraph about being passionate about vehicles. Same trust badges. Same heroic photograph of a spanner.

It is not bad. It is average, expertly rendered. Average is a genuine problem when your entire job online is to give someone a reason to pick you over the four other results on the page.

If your website was generated from a description of your industry, it will read like one. Standing out requires knowing what everyone else already says, and then deliberately not saying it.

The website is not the thing you are buying

This is where most people get their money's worth wrong.

When someone pays for a website, they think they are buying a website. They are not. They are buying enquiries. The website is the delivery mechanism, and it is an easy thing to point at, so it becomes the thing everyone prices.

Judge it that way and the maths goes sideways. A free site that produces nothing is not cheaper than a paid site that brings you two jobs a month. It is a smaller amount of money spent on nothing at all.

I have watched businesses launch a beautiful new site, feel tremendous about it for a fortnight, then quietly notice the phone is doing what it always did. Nothing broke. The site was simply never asked to do the useful thing.

Where AI genuinely earns its keep

I am not anti any of this. We use AI constantly, and the platforms we build on use it too. It writes first drafts and handles the fiddly repetitive bits nobody enjoys. It gets a structure on screen in minutes instead of days, which means more of the budget goes into the thinking and less into pushing boxes around a screen.

That is a good trade. The tools are excellent at execution. They are not good at judgement, and judgement decides whether the thing works.

If you are right at the start, no site, no budget, no customers yet, then absolutely use an AI builder. Existing beats perfect, and it beats nothing by a country mile. Just do not confuse having a website with having one that works, and do not be surprised in six months when it is sitting there being fine and not much else.

The honest test

Open your site on your phone. Ten seconds, the way a stranger would. No scrolling, no cheating, no pretending you do not already know where the contact button lives. Can you tell what the business does, who it is for, and what to do next? Not is it on there somewhere. Can you tell, in ten seconds?

Then the mean one. Does anything on that page give someone a reason to choose you specifically, rather than confirming you exist? Most sites clear the first bar and faceplant on the second. The second one is the one that pays.

The test nobody is running yet

Now the other way round. Open ChatGPT or Claude and ask who they would recommend for what you do in your town. See whether you come up. Then ask what your business does, and see whether the answer is one you would have given.

This is where it gets uncomfortable, because AI-built sites are often the worst performers here, for the same reason they were quick to build. There is nothing specific in them for a model to hold onto, no reason to name you rather than the category. Assembled from the average, filed with the average.

If the answers make you wince, good. It means the gap is not in your design. It is in the thinking underneath, and that is fixable without starting again.

Nobody needs a new website as often as they think they do. Usually the site is fine and the thinking underneath it never happened. If the two tests above turned up something uncomfortable, the fix is probably not a rebuild.