In recent years, several ERP components have been identified as potential neural correlates of consciousness (NCC), including early negativities and late positivities. Based on experiments in the visual modality, it has recently been shown that awareness is often confounded with reporting it, possibly overestimating the NCC. It is unknown whether similar constraints also exist in the auditory modality. In order to address this gap, we presented spoken words in a sustained inattentional deafness paradigm. Electrophysiological responses were obtained in three physically identical experimental conditions that differed only with respect to the participants' instructions. Participants were either left uninformed or informed about the presence of spoken words while confronted with an auditory distractor task (U/I condition), informed about the words while exposed to the same task as before (I condition), or requested to respond to the now task-relevant speech stimuli (TR condition). After completion of the U/I condition, only informed participants reported awareness of the words. In ERPs, awareness of words in the U/I and I condition was accompanied by an anterior auditory awareness negativity (AAN). Only when stimuli were task-relevant, i.e., during the TR condition, late positivities emerged. Taken together, these results indicate that early negativities but not late positivities index awareness across sensory modalities. Thus, they provide evidence for a recurrent processing framework, which highlights the importance of early sensory processing in conscious perception.