Why AI-built websites look great and still lose customers

Why AI-built sites look great

Give the tools their due. AI website builders handle a lot well:

• Visual consistency. The same spacing, colours and component styles from page to page.
• Familiar patterns. Navigation bars, cards, pricing tables and forms that look like the ones people already know.
• Basic responsiveness. Layouts that reflow on smaller screens.
• Speed. What took a developer weeks now takes an afternoon.

That's a genuinely useful starting point. For a lot of small businesses it's a better starting point than a dated template. If you're weighing up where AI fits, I've written about when AI design tools actually make sense and where they don't.

Why they still lose customers

AI generates from patterns, not from understanding your customers. It produces what a website like yours usually looks like. It can't watch someone hesitate on your pricing page, notice that visitors keep scrolling past your call to action because they're looking for proof first, or know that your customers get nervous at step three of the quote form.

So the founder's instinct, to make it look better with more polish and smoother animations, rarely helps. The problem was never visual. It's in six places that AI-built sites get wrong again and again.

1. Forms that work on paper and nowhere else

This is where the gap shows most clearly. AI builds forms with every field you asked for. It doesn't ask whether you need all of them, what order they should go in, how they should be validated, or what happens when someone makes a mistake.

The form looks fine. The completion rate isn't. Common issues: too many required fields, a phone number asked for before you've earned it, error messages that only appear after submit, a cleared form after one mistake, and no confirmation that anything was sent.

2. Empty, error and loading states

AI designs the happy path beautifully: the dashboard full of data, the search with perfect results, the payment that goes through first time. But what does someone see when there's nothing to show yet? When the search finds nothing? When the payment fails or the session expires?

These edge cases are exactly where people need help, and they're often missing or left as a blank screen, a spinner that never stops, or a raw error code. Those moments decide whether people trust you.

3. Mobile that technically works

AI-built layouts usually reflow on a phone, which isn't the same as working well on one. Look for tap targets too small or too close together, sticky banners covering the content, menus that hide the one thing people came for, and forms that open the wrong keyboard. Most visitors to a small business site are on their phones, so this is where the site is judged first.

4. A path to buy that isn't clear

AI creates pages. It's far weaker at building navigation that matches how your customers think about what you sell. The hierarchy makes sense to the tool, not to the person trying to work out which option suits them.

The path from landing to enquiry or checkout also tends to be generic: several competing calls to action, pricing shown before the value is clear, or no obvious next step at the bottom of a long page. Each one costs you a share of the people who were ready to act.

5. Accessibility that only looks accessible

AI-generated designs often look accessible, with decent contrast and readable fonts. Real accessibility goes deeper: visible focus states, keyboard navigation, labels a screen reader can announce, and errors that are announced rather than just turned red. Looking accessible and being accessible are different things. It matters commercially too: these issues block real customers, and in Australia the Disability Discrimination Act applies to websites as well.

6. Missing trust signals

For anything beyond an impulse purchase, people build confidence step by step. AI doesn't know which proof matters to your customers or where it should sit. Reviews and guarantees often land on the homepage and disappear from the checkout or enquiry form, which is where the doubt actually peaks. Placeholder copy, stock testimonials, missing business details, a vague About page and no clear contact information quietly tell visitors you might not be real.

Vibe coding is genuinely useful for getting a site live fast. These six gaps are the cost of stopping there.

A 30-minute self-check for your AI-built site

Grab your phone and someone who has never seen your site, then work through this list:

• Can a first-time visitor say what you do, who it's for and what to do next within five seconds?
• Is there one clear primary action on each key page, or several competing ones?
• Fill in every form on your phone. How many fields could you remove? What happens if you make a mistake?
• Submit an invalid email, leave a required field empty and switch to flight mode mid-submit. Is every message clear and helpful?
• Visit pages with no content yet (an empty cart, a search with no results). Is there something useful there?
• Tab through the site using only the keyboard. Can you always see where you are and reach every link and button?
• Is your proof (reviews, results, guarantees, real business details) visible at the moment someone decides to buy or enquire?
• Check your analytics: which page has the highest exit rate, and which form is started most but finished least?
• Watch five session recordings, if you have them. Where do people hesitate, rage-click or give up?

If you found more than a couple of problems, that's normal. The point of the check is to see where to look first.

When to get a UX audit

You can fix plenty of this yourself. Get outside help when:

• Traffic is reasonable but enquiries or sales aren't, and you can't see why.
• You're about to spend money on ads, which only amplifies whatever is broken.
• You've tried tweaking the design and the numbers haven't moved.
• You're planning a bigger rebuild and want to know what to fix before you start.

You're too close to your own site to see what's broken. A UX audit gives you a ranked list of what's costing you customers on your AI-built site, with a specific fix for each one, so you or your AI tool can make the right changes in the right order.

The bottom line

The AI-built site isn't the problem. The missing UX thinking is. AI got you a site that looks credible in an afternoon. Making it work for the people who land on it (clear forms, helpful states, a usable mobile experience, an obvious path to buy, real accessibility and proof where it counts) is what turns that site into customers.

If you'd like a second pair of eyes on yours, see how my UX audits work. Every audit is a fixed price, quoted upfront after a free 30-minute call.