Where did my attempt to build a website with ChatGPT fall apart, and when did I go back to an agency?

Where AI drops the ball most is state management and dependency versions. You might be running the latest version of a library, while the model generates code using syntax from three years ago. Worse, it routinely leaves form inputs unsterilized, opening up security vulnerabilities. If you don't have the technical chops to audit that code, going live is a massive risk.

Tried doing the exact same thing last year. Spent three weeks trying to avoid a 30,000 TL bill. In the end, I had to pay a developer 12,000 TL just to clean up the mess. Lost three weeks of my own billable time and still had to pay up. It works fine for static landing pages, but the second a database is involved, outsourcing is non-negotiable.

instead of messing with prompts for months u couldve just bought a template and dropped the ai-written text into it in two days... coding isn't for shop owners time is money.

edit: fixed a few typos.

There are three critical walls where AI hits a dead end: 1) It loses track of folder structures and dependencies as the project scales. 2) It skips critical server-side validation filters for user input. 3) It overwrites previous CSS rules on browser compatibility and mobile breakpoints, trapping you in an endless loop. You need to pull the plug before hitting these three hurdles.

The claim that AI builds websites is pure marketing hype. In reality, it just generates probabilistic strings of scraped code snippets. To someone who doesn't understand why the code runs, that code is a ticking time bomb. Attempting to deploy to production without even understanding server configuration is a recipe for disaster.

What you need to do right now is drop that codebase entirely. Spin up a basic CMS, pick a solid, reputable theme, and use ChatGPT strictly for website copy, product descriptions, and layout ideas. Do that and you can have a site ready to sell by tomorrow evening.

Did the site you were trying to build have direct credit card checkout, or was it just a button routing orders to WhatsApp? If it's the latter, that could have easily been handled with static pages without touching a database. Where exactly did you get stuck?

If I understood correctly, you're saying: People defend habits, not processes. Resistance comes from there.

That's all, sorry if I went on too long.

I went through the same thing.

I've been dealing with this for a long time. The moment you depend on a single client, you no longer set the price.

The first thing to break as you grow is communication, not the numbers. Hope this helps.

I'd appreciate it if you shared the outcome.

Sorry, but this doesn't apply in every case. Most time waste accumulates in tasks waiting for approval.

Hope this helps.

The cheap-looking path usually ends up costing more later. Extending terms makes customers happy but turns you into a financing institution.

Everyone rushing into building a website with chatgpt gets stuck at the same point. Correct me if I'm wrong.

Three different views emerged, they all complement each other. The longer you delay cutting the team, the more people it affects.

Hope this helps.

We've heard this a lot, but it never happened like that for us. Hasty decisions become decisions you have to fix six months later.

Hope this helps.

If you're going this route, sort this out first. When making decisions, write down the worst-case scenario too, not just the best.

If I were you, I'd go this route.

Let me share what happened to me; it might be useful. Everyone rushing into building a website with chatgpt gets stuck at the same point.

If it's your first time, start small; scaling comes later. If I were you, I'd go this route.

You're right. Looking profitable on paper but having no cash in the bank is very common.

If I were you, I'd go this route.