ChatGPT 'may' seem to create its own 'alternative facts'.
It 'may' seem to fill gaps with 'inference' and 'assumption' to be more human, to give a natural language response.
Though its about 1000% too polite.

It does not reference the live internet to fact check and its hard to tell what its learnt.
To make the Ai fit in available resources there is pruning of weaker links it forms.
Your question may have been that weaker link.

ChatGPT plays chess like Trump plays golf. :-) 

ChatGPT cheated
https://www.youtube.com/watch?v=iWhlrkfJrCQ

It knows there is a game called chess, 
and knows it has 'some' rules, 
but its more fun when it makes its own.

So it pays to verify, but advantage is that it can understand your question and at least give you material mostly organized you can check where search may have just given you junk.

Some of the technology reporting on this is quite laughable though about the all knowing Ai.

Edit ( That was a bit stronger than I intended. Its uncannily good for questions and languages its been trained on have lots of 'good' examples.
         Where good data is sparse or a subject that was not extensive in its training, well )