The paper verifies the hypothesis of time-dependent dynamics of steady-state visual evoked potentials during a short series of stimulations (15 s) simulating work with brain-computer interfaces. Using deep machine learning of a neural network with direct propagation and known methods of machine classification, the frequency characteristics of the visual evoked potentials of electroencephalography during the work with brain-computer interfaces are analyzed. It is shown that the temporal dynamics of steady-state visual evoked potentials even for such a short period of time can sufficiently change the parameters. It can potentially serve as an obstacle for work with this type of brain-computer interfaces for a number of users. The described approach allows us to confirm the hypothesis that over time the brain shows signs of fatigue, consisting in changes in the frequency-time characteristics of the registered signal. Thus, the human brain shows signs of fatigue during sessions of steady-state visual evoked potentials.