The purpose of the study is to identify the possibilities and limitations of artificial intelligence technologies in political forecasting. The principles of simplified factor analysis, as well as critical discourse analysis of academic literature devoted to the use of AI in political forecasting, served as methodological tools. The analysis showed the inconsistency of the introduction of intelligent systems in the field of modern political forecasting. Firstly, among the trigger factors contributing to the development of political AI prognostics, the following were identified: the ability to process huge amounts of information about politics using AI; working out the reliability of future decisions in the field of domestic and foreign policy; high predictive potentials of neural networks with high-quality organization of training and testing; modeling the political behavior of individuals and social groups; identification of precursor of political events by neural networks. Secondly, the stop factors hindering the development of political AI forecasting include: the effect of «poisoning data» due to its inaccuracy, marginality and unreliability; value bias of intelligent systems; substitution of political forecasting by over-fitting data; imperfection of existing political theories used in conjunction with machine learning; tendency of AI to escalation political modeling. The conclusions indicate that the most important trigger factors include the search for hidden political patterns, as well as the hybridization of machine and traditional political forecasting capabilities. It is emphasized that the prospects of the factor of searching for hidden political patterns can balance and exceed the serious risks of the «black box» effect only in combination with the factor of the possibility of hybridization, corresponding to the parameter of the combining the potentials of all three types of forecasting - trend (mathematical), analytical modeling and expertise.