In recent years artificial intelligence technologies, including machine translation, have been actively developing. Modern neural network systems based on Transformer provide high-accuracy translation of scientific and technical texts, but not in the case of literary translation. Literary text requires the transmission of style, emotional nuances and cultural associations, being difficult for machine systems. The aim is to determine the capabilities and limitations of Machine Translation working with literary texts. The object of analysis is machine and author’s translations of F. Thompson’s novel “Lark Rise to Candleford” (1945). The subject of the analysis is the lexical, morphological, syntactic and pragmatic features of translations. It was done on the base of comparative and component analytical method, as well as a translation experiment involving senior students from the RUDN University Faculty of Philology. The scientific novelty of the work lies in the analysis of a text that has not previously been translated into Russian, so the influence of former versions on the work of machine translation is eliminated. The practical significance of the research is aimed at improving neural network translation models and training specialists in post-editing texts created using AI.