Why I Stopped Trusting AI Answers in College Without Checking Them

When I started using AI for college projects, it felt like a shortcut.

I could ask for code, copy the answer, and finish tasks faster. The problem came when I had to explain the application to my professor.

Sometimes the code worked, but I couldn't clearly explain why a class existed, how the request moved through the system, or what each layer was doing.

For example, I might have had this structure:

Controller
→ Service
→ Repository
→ Database

But when someone asked me to explain the flow, I struggled.

That made me realise that working code is not the same as understanding code.

I stopped copying everything

AI answers can look correct and professional, but they can still include unnecessary patterns, outdated libraries, or incorrect assumptions.

Instead of copying the full answer, I started focusing on the concepts. Now I ask questions such as:

• Why is this class needed?
• What problem does this pattern solve?
• Where does the request start and finish?
• Could this code be simpler?
• Can I explain it without reading it?

I also started treating AI-generated code like a code review. I read the changes, check the dependencies, run the code, and try to understand every important step before accepting it.

Understanding the flow

For a basic .NET API, I should be able to explain:

The client sends a request.
The Controller receives it.
The Service applies the business logic.
The database is updated.
The API returns a response.

If I can't explain that flow, I probably don't understand the application yet.

I still use AI, but now I use it to explain concepts, review code, and suggest improvements.

The main lesson was simple:

AI should support my thinking, not replace it.