Pulsed electrochemical operation can continuously reshape catalyst surfaces, but the atomistic pathways connecting redox cycling to catalytically active structures remain poorly understood. We develop a parallel-evolution grand canonical Monte Carlo/molecular dynamics (PE-GCMC/MD) framework combined with a constant-potential machine-learning force field to simulate Cu-water interfaces under dynamic potentials. Oxidation promotes oxygen incorporation and lattice expansion, whereas reduction triggers progressive deoxygenation and lattice collapse, generating transient low-coordination, pit-like Cu motifs. These non-equilibrium motifs substantially lower the barrier for C-C coupling. By varying pulse amplitude and duration, their populations can be regulated, demonstrating that catalytically active states are encoded by electrochemical history rather than a single equilibrium structure.