# MathScinet reviewing: should I check the proofs?

The guidelines do not clearly specify how much work one is assumed to do in writing a MathSciNet review. What is the consensus? Is one expected to read the whole paper and check the the proofs in detail?