The one difference I can see is how chatGPT optimises the code based on the instruction you provide, while Copilot could only complete the code in one way.
For instance I asked chatGPT to write a code for square root which it did using inbuilt Math.sqrt function. I instructed it to write without using inbuilt function which it did. However, no matter how many times I instructed it to avoid using Math.floor or Math.ceil, it was not taking any heed.
in most copilot editor extentions there is a window you can open that gives 10 different completions. You can also easily shape what it gives you with a single comment ("# function for a qsquare root, don't use math.sqrt or math.floor" etc)
