Building a Website with Claude
I've worked in digital since 2000, and for most of that time I've had the same dream: describe a website in plain language and watch it get built. No specs lost in translation, no waiting on a dev queue. Just say what you want and see it. When this site went live, I told a colleague exactly that: this is what I'd been dreaming about since I started.
I've also spent years on the agency side managing the teams that build websites, so I know what goes into one: design, copy, development, hosting, forms, domains, and a long list of details nobody thinks about until launch day. So when I needed a site for SpotSeven Intelligence, I ran an experiment. Could I finally build the whole thing myself, with Claude as my production partner, and end up with something I'd actually be proud to send to a CMO?
The site you're reading this on is the answer. It took a few days, and I didn't write the code myself. Claude Code did. Here's how it went.
For the record, I'm on a paid Claude Pro plan and ran this experiment in Claude Cowork.
The steps
- Start with what you already have. Before asking for a single line of code, I gave Claude my positioning, services, bio and client list. Placeholder copy is always wrong in ways that matter, and you're the one who has to defend every word on a sales call.
- Describe a direction, not a template. I said "slick and professional, data-driven, nothing that looks basic," and picked a palette: charcoal and soft pink. Then I let Claude propose and I reacted.
- Keep it to one page. The whole site started as a single HTML file. No website builder, no plugins, nothing to update or break.
- React, don't spec. Look at it, ask for one change, look again. Most of my best decisions came from seeing something on screen, not from planning it up front. And if a change makes it worse, ask for the old version back. I did, twice.
- Make it readable by people and by AI. If someone asks ChatGPT or Claude for a marketing ops partner, I want my site in the answer. So the site openly welcomes AI crawlers and includes structured data and a plain-language summary written for them.
- Wire up the working parts. The contact form runs on Netlify Forms. A "Book 20 minutes with Katie" button goes to Cal.com, so people can skip the email back-and-forth.
- Go live. Hosting is on Netlify. The domain is registered at Namecheap, and I only changed the records pointing to the site, so my email kept working.
Lab findings
Claude's knowledge of other products' screens can be out of date. Its instructions for setting up Cal.com described menus that had since been renamed and moved. The fix was simple: send a screenshot of what you actually see, and trust it over the instructions.
The defaults will get you. Netlify turns form detection off on new projects and makes new sites private. Both cost me time on launch day. I had Claude write every one of these into a step-by-step runbook, so the next deploy takes minutes instead of an afternoon.
The best idea I didn't build was a chatbot. My first instinct was an AI assistant on the homepage to talk with visitors. Claude walked me through the tradeoffs: it needs a server, costs money every conversation, invites abuse, and hides your content from the very AI search tools you want to be found by. Instead, visitors pick the situation that sounds like theirs, and the page reorders itself to lead with the services that fit. Same personal feel, none of the overhead.
What I'd tell you
You don't need to be a developer to do this. You do need to know what good looks like, and the patience to look, react and look again. That part is still human work. It also helps to have a working knowledge of how websites are built and everything that goes into them. Otherwise it's easy to miss something important, like SEO.
Twenty-six years is a long time to wait for a dream. It was worth it.
And if any of this sounds like more than you want to take on, reach out. I'd be happy to help, whether that's getting you started or building it with you.