This paper proposes a technique called fuzzing to automatically generate random DRAT refutations and check them against SAT solvers and verified checkers in order to find bugs. The technique found bugs in two existing DRAT checkers, drat-trim and proofcheck, showing they incorrectly reject or accept certain refutations. The paper's contributions are a fuzzing technique to find bugs in DRAT checkers, discovering specific bugs in two checkers, and developing a verified checker in Coq that is complete.