I Asked Claude to Build a Therapy Website. Then I Redesigned It.

AI can build websites now.

At least, that’s what everyone keeps saying.

So I decided to test it myself.

I gave Claude a detailed prompt and asked it to create a website for a fictional therapy practice. Within minutes, it generated the HTML, page structure, headlines, and content sections.

The result was surprisingly good.

But it also highlighted something important:

There’s a difference between generating a website and designing one.

Starting With a Brand, Not a Prompt

For this experiment, I created a concept for a fictional therapy practice.

Since it was only a concept, I didn’t spend weeks developing the brand. But I still started where I start every website project: with the overall feeling.

I named the practice Rhodolite Therapy (yes, I name my concepts after gemstones).

The goal was to create a brand that felt:

• Warm
• Cozy
• Safe
• Trustworthy
• Welcoming

I imagined a therapy practice serving individuals, couples, and teens, with a particular appeal to women.

The visual direction was built around soft pink tones paired with muted olive green. I wanted visitors to feel comfortable and supported from the moment they landed on the website.

Because that’s what branding really is.

It’s not just colors and fonts.

It’s the feeling people get when they interact with your business.

Using Claude as a Website Assistant

Once I had the brand direction, I turned to Claude.

I gave it a prompt describing the therapy practice and asked it to create homepage copy and HTML sections.

The result was surprisingly solid.

Claude generated:

• A homepage structure
• Headlines
• Supporting copy
• Calls to action
• HTML and CSS

From a structural perspective, it did a good job.

In fact, this is exactly how I think AI is most useful for web designers.

Not as a replacement.

As an assistant.

It can help generate ideas, suggest structures, and speed up the first draft process.

What Was Missing

When I looked at the AI-generated website, nothing was technically wrong.

The sections made sense.

The layout worked.

The copy was acceptable.

But it felt generic.

Even if I swapped the images and adjusted a few colors, it still felt like a template generated by a machine.

There was no personality.

No emotion.

No sense of who this therapy practice was.

And that’s where I think a lot of AI-generated websites fall short.

They provide information.

But they don’t create an experience.

Turning the Concept Into a Real Website

I took the same brand, the same content (mostly — I tweaked the copy a bit), and rebuilt everything in Showit. I added pages: About, Services, FAQ, Blog, Contact, Book a Consult. I brought in the pink and olive palette from the mood board, real photography with faces, a subtle text animation on the hero, a video background on a couple of sections to add some life and movement.

The intro section — I call it Mist — was really important to me. You have about three seconds before someone decides whether a website is for them. That first section has to tell the whole story immediately. So I made sure it landed.

Side by side, the difference is obvious. The AI version is a website. The Showit version feels like something.

So, do you still need a designer?

Honestly, it depends on the niche. For B2B, finance, crypto — AI probably works great. The logic-heavy, less emotional stuff? Go for it.

But for anything where a visitor needs to feel something the moment they land — therapy, wellness, personal brands — you need more than structure. You need strategic thinking about what a person experiences on that page. What makes them feel safe. What makes them stay.

I use AI all the time in my process. It’s genuinely useful for building out structure, drafting copy, getting a starting point. But as a helper — not as the whole thing.

A website is emotion and logic. AI is really good at one of those.