When confronted with continuous tasks, humans show spontaneous fluctuations in performance, putatively caused by varying attentional resources allocated to process external information. If neural resources are used to process other, presumably “internal” information, sensory input can be missed and explain an apparent dichotomy of “internal” versus “external” attention. Each of these opposing attentional modes might have their own distinct neural signature. For instance, α-oscillations (∼10-Hz) have been linked to a suppression of sensory information and might therefore reflect a state of internal attention. Indeed, in auditory cortex of macaque monkeys, periods with strong α-oscillations and reduced responses to sounds alternate with periods in which the opposite pattern occurs. In the current study, we extracted neural signatures of internal and external attention in human electroencephalography (EEG) - α-oscillations and neural entrainment, respectively. We then tested whether they exhibit structured fluctuations over time, when listeners attended to an ecologically relevant stimulus like speech, and completed a task that required full and continuous attention. Results showed an antagonistic relation between spontaneous α-oscillations and neural activity synchronized to speech. These opposing neural modes underwent slow, periodic fluctuations around ∼0.07 Hz that were strikingly similar to those observed in non-human primates, and related to the successful detection of auditory targets. Our study might have tapped into a general attentional mechanism that is conserved across species and has important implications for situations in which sustained attention to sensory information is critical. Significance Understanding fluctuations in sustained attention is crucial for a wide variety of everyday activities, e.g., in traffic, education or safety monitoring. Lapses in attention can negatively affect outcomes in these tasks, sometimes with severe consequences. Here we demonstrate that auditory sustained attention fluctuates between an external mode favoring the processing of sensory information and an internal mode, where sensory information is ignored. These attentional modes have their own neural signatures and alternate with an inherent rhythmicity that can be traced in the human electro-encephalogram. Our results may be leveraged to predict attentional fluctuations, optimize critical system designs or tailor interventional approaches to improve sustained attention.