There is a conversation I have been having more frequently. A prospective client, usually somewhere in the early stages of enquiry, mentions my pricing. Then comes the line — delivered with the quiet confidence of someone who has done their research:

“Your prices are expensive. AI can even build websites now.”

This makes me smile.

Not because the statement is wrong. It is not wrong. AI can build websites. AI can also generate logos, write copy, design social media graphics, compose music, and produce passable code. The tools are real, they are improving, and anyone telling you otherwise is not paying attention.

I only smile because the statement is incomplete. And the gap between what it says and what it means is exactly where the problem lives.

What AI Actually Does When It “Builds” a Website

When you prompt an AI tool to build you a website, what you receive is a plausible starting point. It will have structure, some visual hierarchy, probably some default styling, and content that sounds roughly right. If the tool is good and your prompt is specific, it might look quite decent at first glance.

Then you look closer.

The font pairing is generic. The spacing feels off in ways you cannot quite name. The hero section says something that is approximately true about your business but does not actually sound like you. The colour choices are fine but forgettable. The mobile layout has something awkward happening around the third section. The contact form does not connect to anything. The images are placeholders or AI-generated visuals that do not represent your actual product or team.

Now what?

This is where the “AI can build websites” argument quietly collapses — not at the generation stage, but at the modification stage. Because everything AI produces requires judgment to evaluate, and judgment to improve. If you do not have a background in design, you do not have the reference point to know what is wrong, much less how to fix it. You are not stuck with a bad website. You are stuck with a website that looks almost right and feels slightly off and you cannot articulate why, and neither can the AI when you ask it to fix it.

The Graphic Design Parallel

This is not unique to web design. The same dynamic is playing out with graphics.

“I can just prompt a design.” Yes. And what you will get is something that exists at the intersection of what the AI has seen before and what your prompt describes. It will not be wrong. It will not be yours, either. It will lack the intentional decisions — the deliberate type choice, the considered colour weight, the spatial relationship between elements — that make a design feel considered rather than generated.

Designers are not valuable because they operate software. They are valuable because they make decisions. The software is just how the decisions get expressed. AI can operate the software. It cannot make the decisions, because decisions require judgment, and judgment requires taste, which is built over years of looking, making, failing, and looking again.

When a designer with ten years of experience prompts an AI tool, the output is different from when someone with no design background prompts the same tool. Same AI. Same general request. The difference is in what the designer recognizes, what they push back on, what they modify, and what they know to throw away entirely.

The Legitimate Use Case

I want to be clear: there is a version of this that makes sense.

If you run a small operation — a freelancer, a side project, a community group — and your primary need is just an online presence that says “we exist, here is how to reach us,” and you have the time and curiosity to figure it out, then an AI-assisted website might genuinely serve you. The tools have lowered the floor. That is real and it matters for people who previously had no access to a web presence at all.

The question is not whether AI can produce a website. It is whether the website AI produces is the one you actually want — or simply the one you ended up with.

Where It Always Ends Up

Here is what I have observed consistently: clients who build their own AI-assisted sites often come back. Not immediately. Usually six months to a year later, when one of the following has happened:

The site is not showing up in search. Or it is slow and they cannot figure out why. Or they tried to add a new section and broke the layout. Or they rebranded and now nothing looks consistent. Or they want to connect a booking system and hit a wall. Or they simply got busy — which is what happens when you run a business — and the site has been sitting unmaintained for months while the world moved on.

Websites are not finished products. They are living infrastructure. They break. They need updates. They need to evolve as the business evolves. A site built by someone with no web background is not just a site with aesthetic limitations — it is a site with no owner who can maintain it. The moment something goes wrong, you are either paying a professional to fix someone else’s questionable decisions, or you are starting over.

Unless, of course, you intend to become a hobbyist web designer yourself. That is a real choice. Some people make it and find they enjoy it. If that is genuinely what you want to invest your time in, I think you should do it. The tools are accessible enough now that it is not an unreasonable path.

But if you are running a business, your time has a cost. Every hour you spend wrestling with a website builder or debugging an AI-generated layout is an hour you are not spending on the thing you are actually good at. The maths on “free” is rarely as clean as it looks.

What a Professional Actually Brings

When I build a website, the AI tools are part of the workflow. Not hidden, used. For copy drafting, code assistance, image generation for concepts, documentation. The tools are useful and I use them.

The difference is everything that happens around the tools: the strategic thinking about what the site actually needs to do, the performance decisions baked into the build, the security configuration, the accessibility considerations, the decisions about what to include and — equally important — what to leave out.

And when something breaks six months from now, there is someone to call.

That is not something you can prompt.

So when I hear “AI can even build websites,” I smile — because I know the next chapter of that story. It usually ends with a conversation about what it would cost to fix it, or build it properly the second time.

I would rather have that conversation at the beginning.