Recently it has been discovered that visuospatial attention operates rhythmically, rather than being stably employed over time. A low-frequency 7–8 Hz rhythmic mechanism coordinates periodic windows to sample relevant locations and to shift towards other, less relevant locations in a visual scene. Rhythmic sampling theories would predict that when two locations are relevant 8 Hz sampling mechanisms split into two, effectively resulting in a 4 Hz sampling frequency at each location. Therefore, it is expected that rhythmic sampling is influenced by the relative importance of locations for the task at hand. To test this, we employed an orienting task with an arrow cue, where participants were asked to respond to a target presented in one visual field. The cue-to-target interval was systematically varied, allowing us to assess whether performance follows a rhythmic pattern across cue-to-target delays. We manipulated a location’s task relevance by altering the validity of the cue, thereby predicting the correct location in 60%, 80% or 100% of trials. Results revealed significant 4 Hz performance fluctuations at cued right visual field targets with low cue validity (60%), suggesting regular sampling of both locations. With high cue validity (80%), we observed a peak at 8 Hz towards non-cued targets, although not significant. These results were in line with our hypothesis suggesting a goal-directed balancing of attentional sampling (cued location) and shifting (non-cued location) depending on the relevance of locations in a visual scene. However, considering the hemifield specificity of the effect together with the absence of expected effects for cued trials in the high valid conditions we further discuss the interpretation of the data.