Nowadays artificial intelligence finds its application in various branches of economic activity and intellectual sphere. The introduction of neural networks in many areas of human life, determines the relevance of this research. In this paper, the purpose of the research is to study the potential of neural networks and the possibility of their application in conceptual design. The relevance and purpose of the research allowed us to formulate the following tasks, which allow us to evaluate the neural network performance at all stages of conceptual design: analyzing and selecting a number of neural networks generating an image; testing the selected neural networks and approbation of them for a specific image task; final selection of a neural network to perform the design work; the mechanism of working with the neural network to obtain the desired result. The outcome of the article is the actual application of generative neural networks for their integration into architectural projects at the stage of creating conceptual solutions—from idea presentation to visualizations of the object and architectural environment. The realization of the empirical part of the research presented in the article was carried out within the framework of work on the concept idea of “Pulkovo Innovative Astronomical Complex”, the authors of which became prize-winners of the Open All-Russian Architectural Creative Competition ArchConcept.