AI-generated, template, or bespoke: which website should a small business actually buy?

I build websites with AI. Most of my day is spent in Claude Code.

So when I tell you that "just get AI to build it" is bad advice for most small businesses, understand what I'm not saying. I'm not saying the tools are bad. I'm saying a tool is only as good as the person deciding what to do with what comes out of it, and that the hard parts of a website were never the typing.

There are three ways to get a website in 2026. Each one costs you something different, and the cost isn't always the invoice.

Option one: let AI build the whole thing

Type a description, get a website. Lovable, v0, Bolt, Framer's AI tools, Squarespace and Wix's built-in generators. They have improved enormously and they are genuinely impressive.

They also get you about 80% of the way there in an afternoon, and then the remaining 20% eats your month.

What the last 20% actually is

It's the parts you can't describe in a prompt because you don't know they exist yet. The mobile layout that breaks at one specific width. The contact form that submits but sends nothing. The image that loads at four megabytes and drags your page speed down. The heading structure that's invisible to search engines. The colour contrast that fails accessibility checks. The checkout that works until someone pays with a card from another country.

None of these announce themselves. Your site looks finished. It works when you test it, because you test it the way you built it. The failures happen to other people, on other devices, and you never hear about them, because someone who couldn't get your form to submit doesn't email to tell you the form doesn't submit.

The maintenance problem

The bigger issue arrives later. Six months in, you need to change something. Add a service, update your prices, fix a bug the tool introduced in an update.

If you can read the code, this is twenty minutes. If you can't, you're back at the prompt, describing the change and hoping the model doesn't break three other things while making it. Sometimes it doesn't. Sometimes it rewrites a component you needed and you don't notice for a week.

You end up in a strange position: you own something you can't maintain and can't fully explain, and the only way to change it is to ask the machine nicely again.

The sameness problem

Every one of these tools was trained on the same corpus of websites, and they have converged on the same answers. Gradient hero. Three feature cards with icons. Testimonial carousel. Pricing table with the middle one highlighted. FAQ accordion. Footer with four columns.

I'm not describing a specific site. I'm describing all of them.

If your business competes on being findable and forgettable, this is fine. If it competes on judgement, taste, or being someone a client trusts with something that matters, a website that looks like every other website is working against you before anyone reads a word.

When it's the right call

Genuinely: if you need something up this week, you have no budget, and the site's only job is to prove you exist and give people your phone number, use one of these tools. A live site that looks generic beats no site by a distance that is not close. Come back to it when the business can afford to.

Option two: buy a template

A well-built template from a designer who knows what they're doing is a real product, and often good value. Someone has already solved your layout, your type scale and your mobile breakpoints, and solved them properly.

The trade-off is different from the AI one, and it's usually misunderstood.

It's slower than you think

The promise is "live by Friday". The reality is that a template arrives full of someone else's content, and replacing it is a slog. Your photos are the wrong shape for the image slots. Your services don't map to the six blocks provided. Your headline is nineteen words and the design wants seven.

So you either rewrite your business to fit the template, or you start modifying the template, at which point you're editing code you didn't write in a system you don't understand. I've watched people spend three weeks on a template that was sold as a weekend job.

The differentiation problem is real but overstated

Yes, other businesses use your template. In practice most of your clients will never see the other ones, so the risk isn't that you get caught. It's subtler.

A template encodes assumptions about what kind of business you are. It has a services section because most businesses have services. It has a testimonial block in a particular position because that's where testimonials usually go.

If your business is shaped like the assumption, this is helpful. If it isn't, you spend your energy fighting the layout. A production studio that assembles a different freelance team for every project doesn't have a "meet the team" page, and every template will offer it one.

When it's the right call

You're a straightforward business in a well-understood category, your competitors' sites all look roughly alike anyway, and what you need is competence rather than distinction. A good template gets you a professional site for a fraction of bespoke, and there is no shame in that.

Option three: bespoke

Bespoke is more expensive and slower to start. It is worth being honest about that rather than pretending otherwise.

What you get for it is a site built around how your business actually works, rather than around how websites usually work.

An example

A creative production studio I built for doesn't have a fixed team. The founder assembles a different bench of freelancers for every project, which means the usual "meet the team" page would actively misrepresent the business.

The contact page is where it mattered most. A standard contact form is a name, an email and an open text box, which produces enquiries like "hi, how much for a website?" and a week of back-and-forth to find out whether the project is even viable.

Instead, the form asks which of nine services the project needs and which budget band it sits in. Enquiries arrive pre-qualified. The studio can tell in ten seconds whether a brief is worth a call.

No template ships with that, and no AI builder will invent it, because neither knows how that business makes decisions. It came out of a conversation.

The part that gets undersold

Bespoke doesn't only mean it looks different. It means someone understands how it's put together and can change it later without guessing.

When your prices change, when a payment provider deprecates an API, when Google changes what it rewards, when you add a service, someone can open the code and make the change. That's not a luxury. It's the difference between a website that grows with the business and one that quietly rots until it gets replaced.

When it's not the right call

If you're pre-revenue and testing whether anyone wants what you sell, don't buy bespoke. Find out first. If your website genuinely only needs to display your opening hours, don't buy bespoke. And if the only thing you want is for it to look nicer, be aware that a nicer-looking site with the same underlying problems is an expensive coat of paint.

So where does AI actually fit?

Back where I started, because the honest answer matters more than a tidy one.

I use AI on every build. It writes the boilerplate, scaffolds the components, handles the repetitive work that used to eat days.

The difference is what happens to the output. I can read what it produces, so when it generates something subtly wrong, which it does, I catch it. When it takes a shortcut that will cause a problem in eight months, I see the shortcut. I know which of its suggestions to throw away, which is most of them, because the default answer it reaches for is the average of every website it has ever seen and your business is not average.

Then there's everything after the code. Where it's hosted. How the domain and email are configured. What happens when a dependency has a security patch, or a payment provider changes an API, or the site needs to be moved. None of that appears in a prompt, and all of it decides whether the site is still working in two years.

The tool didn't replace the judgement. It removed the typing.

That's the actual line, and it isn't AI versus human. It's whether the person holding the tool can tell when it's wrong, and whether anyone is looking after the thing once it's live.

The short version

AI-generated is fast and free and gets you 80% of a website. Perfect if you need to exist online by Thursday. Difficult to maintain if you can't code, and it will look like everything else.

Templates are good value and professionally designed. Slower to populate than advertised, and they'll gently reshape your business to fit their assumptions.

Bespoke costs more and starts slower. You get a site built around how your business actually works, and someone who can change it when the business changes.

Most small businesses should start with one of the first two and move to the third when the website starts costing them money by being wrong. The mistake isn't buying the cheap option. It's staying on it long after it's become the expensive one.