When I decided to build The AI Review Hub, I thought the difficult part would be writing the articles. I genuinely believed creating the website would be easy because AI website builders promised to do most of the work.
I was wrong.
The site looked good—until I started using it
I began with Hostinger AI and asked it to create a simple blog. The first result looked surprisingly good. The layout was clean, the pages appeared professional, and for a moment I felt that my first website was almost ready.
Then I started clicking around.
The website looked complete, but much of the functionality did not work. Even simple tasks such as changing my contact information, connecting my LinkedIn profile, or updating the Terms and Conditions became difficult.
To manage these basic elements, I tried developing a backend administration page. The idea was to have one place where I could update the website without touching the code every time.
Once that was added, I became more ambitious. I introduced user registration so that other people could log in and submit content. I also wanted an administrator approval process, ensuring that nothing could appear on the website until I had reviewed it.
That was the point where the site started falling apart.
The login, content submission and approval workflows did not connect properly. I kept asking the AI to identify and fix the problems. It suggested changes, but it never appeared to find the root cause. Sometimes one issue seemed fixed, only for another one to appear.
Meanwhile, my AI credits kept getting consumed.
As a non-developer, I could not tell whether the problem was in the code, database, hosting configuration or the way the different components were connected. Eventually, I had a good-looking website that I could not operate or trust.
After spending a lot of time trying to rescue it, I deleted the entire site. My first attempt had failed badly.
Paying for Claude did not solve the practical problem
I still wanted to build The AI Review Hub, so I purchased a paid Claude subscription and started again.
Claude gave me genuinely good ideas. Its recommendations around the website structure, features and code were thoughtful. The output often looked technically impressive.
The problem was what happened after the code was generated.
I still needed to understand where the files should go, how they should connect, what needed to be configured and how to publish everything without breaking it. The solution might have been useful for an experienced developer, but it was not feasible for someone like me.
I had plenty of ideas and pieces of code, but still no working website.
ChatGPT initially produced a similar result
I then switched to ChatGPT. Its explanations were simpler and easier to follow, but the initial outcome was still similar to Claude: advice, suggested code and several different steps that I would have to connect myself.
The real change came when I stopped asking AI simply to “build a website” and clearly explained my constraints.
I said that I was not a developer. I did not want scattered code or instructions that assumed technical knowledge. I needed a complete website packaged as a ZIP file that I could upload, along with simple, step-by-step publishing instructions.
Once I added those constraints, the response changed.
ChatGPT provided the website in an uploadable ZIP package and explained how to publish it. Instead of giving me more technical pieces to assemble, it gave me something I could actually use.
That was the first time I moved from an AI-generated concept to a working website.
What I learned from the experience
This is not simply a story about one AI tool being better than another. Someone with development experience might have achieved a very different result with the same tools.
For me, the most important lesson was that AI needs to understand not only what you want, but also what you are capable of implementing.
A beautiful interface is not a working product. Good code is not automatically a usable solution. For a non-developer, the final delivery format, publishing instructions and ability to maintain the website matter as much as the initial design.
The AI Review Hub exists because I failed, deleted everything and tried again with clearer constraints.
That experience will also guide my reviews. I want to look beyond impressive demonstrations and examine whether AI tools deliver practical results for ordinary users—not just for people who already understand the technology.