· 7 min read
Can You Build Your Business Website With AI? Yes. Here's What It Won't Tell You
AI can build a good-looking website in an afternoon. What it can't do is tell you what you forgot to ask for. Here's what we find under the hood of AI-built sites, and when doing it yourself actually makes sense.
More and more projects start with a message like this: "I built the site myself with AI over the weekend. Can you just put it live?"
And usually, it looks good. Clean layout, decent copy, a contact form, maybe a bit of animation on the hero. A few years ago that would have taken a designer the best part of a week. Now someone with no coding background can get there in an afternoon, and that's a real change.
So the question we keep getting, sometimes politely and sometimes not, is fair: if AI can build a website, why pay a developer?
Here's our answer, from a studio that uses these tools every day.
AI is very good at the part you can see
We're not going to pretend otherwise. Tools like ChatGPT, Claude, v0, Lovable and Bolt can produce a working page faster than any person can type. They write reasonable first drafts of copy. They can turn a rough sketch into a layout in minutes. We use them on almost every project, and they've made our own work faster.
The catch is that a business website is mostly the part you can't see. How it looks is the quickest thing to get right. What takes time, and experience, is everything underneath: where the form data goes, who can read your database, how the site behaves on a cheap phone on a bad connection, and whether Google can make sense of it.
A screenshot can't show you any of that. Neither can the AI preview window.
What we usually find when we look under the hood
When clients send us a site they built with AI tools, the same problems come up again and again:
The contact form "works" but the messages never arrive. Or they land in spam, because the domain has no SPF or DKIM records set up. The owner finds out three months later when a customer calls to ask why nobody replied.
API keys sitting in the front-end code. Anyone who opens the browser's developer tools can copy them, and some of those keys are tied to paid accounts.
A database anyone can read. Many AI app builders use Supabase or Firebase. If row level security is switched off, a visitor with a little know-how can pull every customer record. The site still looks perfect while this is happening.
Hero images of 3 or 4MB each, loading in full on every phone, every visit.
Every page with the same title tag, or none at all. No sitemap. Sometimes schema markup with made-up star ratings, which can get a site penalised.
A privacy policy copied from somewhere else that mentions services the business has never used.
No safe way to make changes. Every new prompt fixes one thing and quietly breaks another, and there's no version history to go back to.
None of these are rare. And none of them show up until something goes wrong.
The real skill is knowing what to ask
AI answers the question you give it. It doesn't tell you which questions you forgot. That gap is where experience lives.
Here's a prompt a business owner might write:
Make a booking page for my dental clinic.
Here's roughly what a developer would ask for:
Add a booking form with server-side validation, rate limiting and a honeypot field for spam. Store bookings in Postgres with row level security so patients can only see their own. Send confirmations through our transactional email provider. Keep the form usable with a keyboard and a screen reader, and don't add more than a few KB of JavaScript to the page.
Both get a booking page. Only one gets a booking page you can trust with patient details.
The second half of the skill is reading what comes back. AI is often confidently wrong. It will invent a package that doesn't exist, use a function that was removed two versions ago, or "fix" an error by deleting the check that was catching it. If you can read the code, you spot it in seconds. If you can't, it ships.
Experience is mostly a memory of things breaking
Years in this industry mostly means you've seen a lot of things go wrong. You know that payment webhooks sometimes fire twice and an order gets charged twice if nobody handles it. You know a staging site left on noindex can wipe a business out of Google overnight. You know which cheap hosts slow to a crawl when a post goes viral.
That memory is what turns into a better prompt and a better review. The AI doesn't have it. It doesn't know your business, your traffic, your customers or your legal obligations, and it won't warn you about a problem you never mentioned.
This is also why a developer who uses AI well is faster than both a developer without it and a business owner with it. The tool takes care of the typing. The person takes care of the judgement.
When doing it yourself makes sense
To be fair, not everything needs a professional. Building it yourself with AI is a perfectly good choice for:
A simple one page site for an event, a side project or a personal portfolio
A prototype to show your developer what you have in mind. This one is great. It can save weeks of back and forth.
First drafts of your website copy. You know your business better than anyone, and AI helps get it out of your head and onto the page.
Internal tools that only your team uses and that don't hold customer data
Once a site takes payments, stores personal data, needs to rank on Google or brings in a real share of your revenue, it's worth having someone who has done this before build it, or at the very least review it.
How to get the best of both
The smartest clients we work with don't see this as AI versus developers. They use both.
Bring your AI prototype to the first call. It shows exactly what you like, faster than any brief. We'll tell you what we can keep and what needs rebuilding.
Ask your developer how they use AI. A good one will have a clear answer. If the tools make them faster, that should show up in your timeline.
Pay for a review if you don't need a full build. A few hours of someone checking security, forms, speed and SEO catches most of the problems listed above.
Judge the result, not the hours. The value isn't how long the code took to write. It's that the site works, stays up, keeps your data safe and doesn't need rebuilding next year.
Questions we often get
Can AI replace web developers?
For simple pages, AI tools can now do a lot of what a junior developer used to do. For business sites that take payments, store data or need to rank in search, you still need someone who knows what to ask for and can spot when the output is wrong. In practice, AI has made good developers faster rather than unnecessary.
Is a website built with AI bad for SEO?
Not by itself. Google cares about whether a page is useful, fast and easy to understand, not about which tool wrote the code. Problems come from what's often missing in AI-built sites: unique titles and descriptions, a sitemap, proper headings, internal links, fast images and content that actually says something specific.
Is it safe to launch a site I made with an AI builder?
If it's a simple information page with no logins, forms or payments, usually yes. If it collects any personal data or takes payments, have someone check how the data is stored, who can access it and where your API keys are before you go live.
Should I tell my developer I used AI?
Yes. It's not something to hide. A prototype made with AI is one of the most useful things you can bring to a first meeting.
Built something with AI and want a second opinion before it goes live? Send it over. We'll tell you plainly what's solid, what's risky and what needs fixing. Book a call with Grayshaper.
