In task-switching paradigms, reaction time (RT) switch cost is eliminated on trials after a no-go trial (no-go/go sequence effect). We examined the locus of no-go interference on task-switching performance by comparing the event-related potential (ERP) time course of go/go and no-go/go sequences from cue onset to response execution. We also examined whether noninformative trials (i.e., delayed reconfiguration, no response inhibition) produce similar sequence effects. Participants switched using informative and noninformative cues (Experiment 2) intermixed with no-go trials (Experiment 1). Repeat RT was slower for both no-go/informative (pNG/I) and noninformative/informative (pNI/I) than informative/informative sequences. ERPs linked to anticipatory preparation showed no effect of trial sequence. ERPs indicated that pNG/I sequences reduce response readiness whereas pNI/I sequences reduce repetition benefit for repeat trials. Implications for task-switching models are discussed.