Introduction. Early diagnosis of autism spectrum disorder in children is significant clinical problem due to the ever-increasing incidence of this condition in children. Aim. The aim of the study is a comprehensive analysis of data on the features of the bioelectrical activity of the brain in children with autism spectrum disorders, identification of the electroencephalography significance in the diagnosis of these disorders, including the ability to differentiate their subtypes. Material and Methods. The materials for the review were scientific articles indexed in Russian and international databases (Russian Science Citation Index, Scopus, Pubmed) for the period from 1995 to 2022. Results and discussion. The article presents systematized data related to the analysis of electroencephalograms in children with autism spectrum disorder which is carried out in different ways: visual and quantitative. The authors analyzed the changes in the electroencephalograms obtained using spectral analysis, identifying the functional connections of different areas of the brain, with the implementation of non-linear assessment methods. The theory of «mirror neurons» was discussed in connection with the peculiarities of the sensorimotor rhythm reactivity in children with autism spectrum disorder. Conclusion. The presence of electroencephalographic changes in children with autism spectrum disorders is confirmed, especially in the period from 3 to 12 months, although the degree of the sensitivity of methods for their detection at this stage is insufficient for accurate diagnosis. Dynamic electroencephalographic monitoring should be recommended for children with autistic disorders, especially if subclinical epileptiform activity is detected. It is also worth paying attention to the change in functional rhythms in dynamics. In general, results of electroencephalography are often more informative than neuroimaging methods, which in most cases do not reveal organic brain damage in the presence of obvious developmental disorder in a child. The most promising at present are non-linear methods of electroencephalograms analysis.