My own experience with ChatGPT.

I asked it to write a Bash script to simulate the Knight's Tour (chess). It gave me a minimal, simple simon, 4x4 version. I found this less than impressive. Why? Because, some years back, I had written a full 8x8 implementation, incorporating Warndorff's Algorithm. It was tricky and complex, and it took me months to get it right.

My conclusion is that ChatGPT 3.5 takes shorcuts whenever it can and is not above cheating. I find it of only limited use, at best.