← ALL POSTS
July 28, 2026 5 MIN READ

One prompt gets you a website. Mine took 300

Everyone online is showing off websites built in one prompt. I counted the prompts it took to rebuild webodew.com. It was 300, and I sorted every one into five buckets. Here's the breakdown.

One prompt gets you a website. Mine took 300.
The whole story in 10 minutes. Or read it below.

With all the AI hype, people are being told they can build a great website with one prompt.

I just redid my whole site. So I went back and counted how many prompts it actually took me.

Not one. Three hundred.

Here’s why that number matters, and what all 300 were actually for.

The one-prompt trap

You’ve seen the posts. “One prompt, look what I made.” Some people even share the prompt. You paste it in, and it works. You get a nav, a hero, a few content sections, a footer. Maybe some little cards. A sentence like “we help your business grow.”

It looks fine. That’s the problem.

A one-prompt site is a template. It looks the same as everyone else’s. And a website that looks like everyone else’s isn’t a bad website. It’s a forgettable one. When you sound like everyone else, no one remembers you.

That’s the thing I’m fighting against. Not ugly. Forgettable.

I asked Claude to count

I had Claude go back through our conversations and tally it up. We’d had 55 sessions together. 14 of them were the new website. The rest was thinking through the migration and what to improve.

300 prompts from me. 171 of them landed in a single five-day marathon, concept to live.

None of those prompts said “make a website.” So I sorted all 300 into five buckets: messaging, design, development, feedback, and the invisible SEO work. Here’s what each one was really doing.

Bucket 1: Story, voice and messaging (17)

This is the “something just doesn’t feel good” bucket.

Early on I looked at my hero section and told Claude I was worried we were going too vague. That one nervous prompt kicked off a back-and-forth that landed on three words: clear, memorable, and fun. I love those words. I want my brand known for them. That never happens in one prompt. It happens in the arguing.

Bucket 2: The design system (13)

Before building a single page, I wanted the rules nailed down. Colors. Fonts, from headlines to body. Navigation. Even motion, because I wanted the site to feel alive.

I looked at a few directions. One leaned corporate and professional. I landed on bold, big messages, the way I feel about sports. That one felt like me. It’s also just easier to read.

Bucket 3: The build (155)

About half the prompts were the actual build. This is where AI blows away anything I can do at that speed.

“Build the about page, pull from the existing site, and think about how we make the layout better.” It pumps out something in minutes. I look, I react, I ask for the next round. The traditional way, that loop takes days. This collapses it to minutes.

But notice: all the thinking, the design, the messaging had to come first. The build is fast because the judgment already happened.

Bucket 4: Pushback and refinement (85)

More than one in four prompts was me pushing back.

I’d build something, walk away, come back with fresh eyes and think “that’s off.” The call to action was one. “Schedule a call” felt flat, so I pushed for something more personal. That’s where “Talk to Mike” came from. Old language from the previous site kept sneaking in, and I’d flag it, and Claude, now that it understood the new design system, could spot the mismatch and fix it.

This is the underrated part of working with a machine. It doesn’t mind if you ask it the same thing five ways. You get to test everything, fast, including the annoying stuff. That wasn’t one-prompt magic. It was several prompts, and that’s okay.

Bucket 5: The invisible stuff (30)

SEO and AI visibility. The work nobody sees.

This is my home turf from years of Webflow builds. Proper heading structure. Meta descriptions that aren’t too long. Schema across everything. Little things like author info on blog posts that help AI tools trust the page. 301 redirects so I didn’t torch years of rankings. Person schema. Business schema.

A one-prompt website doesn’t do any of this. I knew the language to ask for it, and it took about 30 prompts to get it right.

So did AI build my website?

Honestly? Yeah. It wrote all the code.

But those 300 prompts are the difference. Concept to live in five days, then a month of quiet refinement as things settled. AI doesn’t replace your judgment. It just makes the tasks faster.

One prompt gets you a website. 300 gets you a site that sounds like you.

See the payoff
What did 300 prompts actually produce?

The clearest proof was speed. Mobile went from a launch-day 64 to a 98, and the biggest paint dropped from 12 seconds to 2. See the 64-to-98 story →

Want to know how yours sounds?

If you’re worried your site sounds like everyone else’s, I do a thing called a Sound Check. I record myself going through your site as a first-time visitor and tell you exactly where the message is getting lost, before you spend anything real on a rebuild.

Get a Sound Check and let’s make your site sound like you.