We compared how different levels of cognitive load affect frontal P3 (fP3) Event-Related Potential (ERP) to novel sounds. Previous studies demonstrated the predictive value of the probe-elicited frontal P3 (fP3) ERP for subsequent detection failures. They also demonstrated how fP3 is reduced when performing visual and/or manual and/or cognitively demanding tasks. These results are consistent with fP3 indexing orienting to novels or, more neutrally: susceptibility. Here, we tested how fP3 is affected by a threefold variation of cognitive load induced by the verb (generation) task. Participants heard a noun and either listened to it, repeated it, or generated a semantically related verb. These conditions were manipulated between groups. One group (N = 16) experienced the listen and repeat condition; the other group (N = 16) experienced the listen and generate condition. When fP3 was probed 0 or 200 ms after noun offset, it was reduced (relative to no noun) only while repeating or generating, not while listening. An additional probe-elicited ERP was identified as novelty-related negativity, and its contaminating influence on fP3 estimation accounted for by a novel vector-filter procedure. We conclude that cognitive load does not affect fP3-indexed susceptibility. Instead, fP3-indexed susceptibility is affected by presentation of the stimulus, with the most pronounced effect in conditions where a vocal response is needed (i.e., repeat or generate, but not listen), independent of the complexity of the response., (Copyright © 2024 The Authors. Published by Elsevier B.V. All rights reserved.)