We investigated the underlying processes that enable improving filtering irrelevant items from entering visual working memory (WM). To this end, participants performed a bilateral change‐detection task in which either targets or targets along with distractors (i.e., the filtering condition) appeared in the memory array while ERPs were recorded. In the cue‐present condition, we provided a spatial cue coupled with a temporal cue regarding where and when the distractors would appear. On some of the filtering trials, after the offset of the memory array, task‐irrelevant probes were briefly flashed either at the locations of the targets or at the locations of the distractors. This enabled measuring whether reactivating the filtering settings resulted in reducing spatial attentional resources to the distractors, allocating additional spatial attentional resources to the targets, or both, as was measured by the P1/N1 amplitude. Results revealed that, relative to the cue‐absent condition, in the cue‐present condition the P1/N1 amplitude was reduced for probes at the distractors and was similar for probes at the targets. In addition, the reduction in the P1/N1 amplitude was accompanied by a reduced filtering cost in accuracy performance in the cue‐present condition relative to the cue‐absent condition. These findings suggest that reactivating the distractor filtering settings improved filtering performance in visual WM by reducing the allocation of spatial attention to the distractors already at early processing stages, and not by allocating additional spatial attentional resources to the targets.