Four Months, Two AI Tools, One Lesson: What ChatGPT and Claude Taught Me About Building a Website

 Four Months, Two AI Tools, One Lesson

For the past four to five months, I've been trying to do something that sounded simple in my head: build a website using AI. I thought it would take a few days, maybe a couple of weeks at most. Instead, it turned into a long stretch of trial and error, confusion, small wins, bigger setbacks, and eventually, two different AI tools before I actually understood what I was doing wrong the whole time.

This isn't a technical tutorial. It's just what actually happened, month by month, and the one piece of advice that finally made things click.

Where I Started

I began with ChatGPT. In my head, the plan was simple — I would describe what I wanted, and it would hand me a finished, working product. That's not exactly how it works, and it took me a while to accept that.

I kept going back and forth. I'd ask for one thing, get a response, realize it wasn't quite what I needed, ask again, get something slightly different, and repeat. Weeks passed like this. I was making small changes here and there — a bit of design, a bit of text, a new page, a different layout — without ever stepping back to ask whether any of it was actually connected to a real plan.

Looking back, I think the biggest problem wasn't the tool. It was that I didn't know what order things needed to happen in. I was treating the whole project like a random list of tasks instead of a process with a clear starting point.

The Middle Stretch — A Lot of Repetition

Somewhere around the second and third month, I noticed I was repeating the same mistakes. I'd fix something, feel good about it for a day, and then realize it created a new problem somewhere else. I was spending more time undoing things than actually building.

During this period, I also tried publishing content on my own, without really having a system. I wrote posts whenever I felt like it, added pages whenever I remembered to, and hoped that eventually things would just start working on their own. They didn't. I didn't understand yet that a website isn't just a collection of pages — it needs a foundation first, and content that actually holds up on its own.

This is the part of the journey I don't think people talk about enough. It's not one dramatic failure. It's a slow build-up of small, repeated mistakes that eventually make you stop and ask, "Why isn't any of this working?"

Switching to Claude

Eventually, I decided to try something different and moved over to Claude. Honestly, the way it explained things felt different from the start. It didn't just hand me instructions and move on — it broke down why something needed to happen in a certain order, and what would go wrong if I skipped a step.

One thing that stuck with me more than anything else: my website's infrastructure — the design, the extra pages, the technical setup, all the things that make a site "look" finished — was already sitting there. But Claude made it clear that this wasn't the priority yet.

It told me plainly: fix your existing content first, then focus on the infrastructure. Don't spend energy making things look polished if the core of the site — the actual posts, the actual value for a reader — isn't solid yet.

That single piece of advice reordered my entire approach. I had spent months trying to make everything look complete before I had even properly checked whether my content was doing what it was supposed to do in the first place. It was like decorating a house before finishing the walls.

Why This Order Actually Makes Sense

Once I understood the logic behind it, it seemed obvious in hindsight. A website with perfect design and empty or weak content doesn't help anyone. A website with genuinely useful content, even with a simple design, has something real to offer. Search engines and readers both respond to substance before polish.

Claude helped me see that I had the priorities backwards. Instead of asking "how do I make this look professional," I needed to be asking "does this actually say anything useful, and is it easy for someone to find and trust." Once the content side was addressed properly, the infrastructure work — the design, the extra pages, the visual polish — would actually mean something, because there would be something worth polishing.

The Real Lesson

Looking back at these four to five months, here's what I'd tell anyone trying to do something similar: AI is powerful, but it works best when you already understand your own priorities, or when the tool you're using helps you build that understanding instead of just following orders.

ChatGPT gave me answers. Claude gave me a process. That difference mattered more than I expected. It didn't just answer my questions — it helped me figure out which questions I should have been asking from the very beginning.

If you're thinking about using AI to build or fix a website, I'd genuinely recommend trying Claude for this kind of work. You can check it out here: 

claude.ai

A Simple Way to Understand AI

Here's something I've come to believe after these several months of going back and forth: understanding how to use AI doesn't require some deep technical background or years of experience. For most people, this much is enough — you ask, you listen carefully to what comes back, and you adjust based on that.

And if you want a simple, almost physical way to really grasp why AI responses can vary so much depending on how you ask, look at the five fingers on your own hand. They're not equal. Different lengths, different shapes, different roles, all attached to the same hand. AI works in a surprisingly similar way. Ask it the same question in a slightly different tone, order, or wording, and you'll get a different answer — shaped differently, useful in its own particular way, but still coming from the same underlying "hand" of technology.

Once you understand that one idea, using AI becomes a lot less confusing. You stop expecting one perfect answer and start understanding that each response is just one finger of a much bigger hand — useful, but only one part of the whole picture.

Where I Am Now

I'm still not finished. The content side of my website is something I'm actively working through, one post at a time, going back and fixing things that were rushed the first time around. But for the first time in months, I feel like I'm working in the right order instead of guessing.

If there's one takeaway from these four to five months, it's this: don't rush to make things look finished before they actually are. Fix the foundation first. The rest will follow, and it will follow a lot more smoothly than trying to do everything at once.