The grammar $A \rightarrow AA \mid (A) \mid \epsilon$ is not suitable for predictive-parsing because the grammar is:
both A and B can be answers but A is a better answer. Because we have standard procedure for removing left-recursion but ambiguity is not easy to remove. - checking if a given CFG is ambiguous is a undecidable problem.
Any Grammar which is Left Recursive will cause any Predictive Parser to fall into an infinite loop. No matter if it is ambiguous or not, it won't be parsable. This is more stronger than saying it is ambiguous so fails.
Hence, answer = option B