It is well established that we can focally attend to a specific region in visual space without shifting our eyes, so as to extract action-relevant sensory information from covertly attended locations. The underlying mechanisms that determine how fast we engage our attentional spotlight in visual-search scenarios, however, remain controversial. One dominant view advocated by perceptual decision-making models holds that the times taken for focal-attentional selection are mediated by an internal template that biases perceptual coding and selection decisions exclusively through target-defining feature coding. This notion directly predicts that search times remain unaffected whether or not participants can anticipate the upcoming distractor context. Here we tested this hypothesis by employing an illusory-figure localization task that required participants to search for an invariant target amongst a variable distractor context, which gradually changed--either randomly or predictably--as a function of distractor-target similarity. We observed a graded decrease in internal focal-attentional selection times--correlated with external behavioral latencies--for distractor contexts of higher relative to lower similarity to the target. Critically, for low but not intermediate and high distractor-target similarity, these context-driven effects were cortically and behaviorally amplified when participants could reliably predict the type of distractors. This interactive pattern demonstrates that search guidance signals can integrate information about distractor, in addition to target, identities to optimize distractor-target competition for focal-attentional selection.