AI-built websites: what you gain and what you risk

What vibe coding is

The term vibe coding was coined by Andrej Karpathy in February 2025, and within months it was Collins Dictionary’s word of the year. The idea is simple: you don’t write code. You describe in plain language what you want, and an AI tool writes the code for you. If you don’t like something, you tell it and it changes it.

Tools like Lovable, Bolt, v0 or Replit build whole websites and apps from a description. The well-known site builders have also added AI that lays out pages on its own. The result often looks impressive within the first hour. The question is what lies beneath the surface.

What you gain: speed, cost, testing

Let’s start with the good part, because it’s real. Vibe coding is excellent when you want to test an idea quickly:

• A prototype to show a partner or investor how something will work, before you spend on the proper build.
• A first draft of your site, to get clear on what you want before you build it properly.
• An internal tool, such as a form or a price calculator, that only your team will use.

In these cases, AI saves days of work and money. And when it’s time for the real site, you’ll know what you want and be able to explain it far more clearly. The catch is that many people look at the prototype, think about what a website costs to build properly, and decide to put it live as it is.

What you risk: security and data

A site that “works” isn’t necessarily secure. Studies from 2025 and 2026 keep reaching the same conclusion: a significant share of the code AI writes has security gaps. Scans of thousands of vibe-coded apps found exposed access keys and users’ personal data that anyone could see.

The reason isn’t that AI is “bad”. It’s that it does exactly what you ask, and you rarely ask for security. If you say “build me a booking form”, it will build a booking form. It won’t ask you who can read other people’s bookings. And when you give an AI agent access to real data, mistakes can be serious: in 2025, a widely reported case saw an agent delete a company’s production database, even though it had been told explicitly not to change anything.

For a prototype, that doesn’t matter much. For a site with forms, customer accounts or payments, basic security measures aren’t optional, and someone needs to check them before it goes live.

What you risk: Google and AI don’t always see your site

Many AI tools produced, and some still produce, sites that “build” their content inside the browser with JavaScript. Visitors see the page normally. Google reads it late or not in full, and several AI crawlers, like the ones ChatGPT uses, barely see it at all. Some tools fixed this during 2026, but it depends on when and how each site was built.

That means you can have a beautiful site that, to search engines, is an almost empty page. If you want ChatGPT to find you, and Google too, this is the first thing to check.

There are simpler issues as well: titles and descriptions that are missing or identical on every page, heavy images that slow loading, buttons that don’t display well on mobile. AI builds something that looks right. It doesn’t check on its own whether it makes the usual UX mistakes that drive customers away.

What you don’t see: accessibility, GDPR, maintenance

There’s also a group of issues that don’t show up in a first impression at all, but have consequences:

• Accessibility: since 28 June 2025, the European Accessibility Act has applied to e-shops and many digital services. A site that not everyone can use may not just be awkward, but out of line with the rules.
• GDPR and cookies: every form, every analytics tool and every pixel comes with obligations. AI rarely sets these up correctly on its own, and the mistakes don’t show until someone asks you for an explanation.
• Maintenance: when something breaks in six months, who understands the code? If the site lives only inside the platform that built it, moving or changing it can cost more than you saved at the start.

None of this is a reason not to use AI. It’s a reason to know what needs checking before an AI-built site becomes your real site.

How we use it

We use AI tools to build websites too, every day. They save us time on repetitive parts, first drafts and testing. The difference is that AI doesn’t make the decisions on its own.

The site’s structure, security, SEO, speed and accessibility are reviewed by a person who knows what to look for. Every site goes through a series of checks before it goes live, whether a person or AI wrote it. What those checks are, and how they adapt to an e-shop, a medical practice or a corporate site, is part of the work we do for each client.

In short: AI makes the work faster. It doesn’t make it less important.

FAQ about AI-built websites

Can I build my business website myself with AI?

You can build something that works. Whether it brings in customers, whether Google finds it and whether it’s within the rules isn’t obvious at a glance. That’s why even a simple brochure site is worth checking before it goes live, and the more data and payments go through it, the more essential that check becomes.

Will Google find me if my site was built with AI?

Google doesn’t “penalise” a site because AI built it. The problem is technical: if the content isn’t on the page when Google reads it, Google can’t find it. That can be checked.

I’ve already built a site with AI. Do I need to rebuild it from scratch?

Not necessarily. A good review and a few fixes are often enough. Whether it’s worth keeping or rebuilding becomes clear once you see what’s underneath.

Is an AI-built site cheaper?

At the start, almost always. Over time, it depends on maintenance, platform subscriptions and what needs fixing later.

Vibe coding is a very powerful tool, as long as you know when it stops being a prototype and becomes your business website. If you already have an AI-built site, or are thinking of building one, at CNCTECH we check it for security, SEO, speed and accessibility, and tell you what’s worth keeping and what to change.

This article describes the tools and rules as they stood in November 2026. AI tools change very quickly. For legal obligations such as accessibility and GDPR, consult a legal adviser.