In this contribution we revisit regular model checking, a powerful framework that has been successfully applied for the verification of infinite-state systems, especially parameterized systems (concurrent systems with an arbitrary number of processes). We provide a reformulation of regular model checking with length-preserving transducers in terms of existential second-order theory over automatic structures. We argue that this is a natural formulation that enables us tap into powerful synthesis techniques that have been extensively studied in the software verification community. More precisely, in this formulation the first-order part represents the verification conditions for the desired correctness property (for which we have complete solvers), whereas the existentially quantified second-order variables represent the relations to be synthesized. We show that many interesting correctness properties can be formulated in this way, examples being safety, liveness, bisimilarity, and games. More importantly, we show that this new formulation allows new interesting benchmarks (and old regular model checking benchmarks that were previously believed to be difficult), especially in the domain of parameterized system verification, to be solved.