The objective of this paper is to propose a novel approach for an intelligent selection of relevant scenarios for the certification of automated vehicles. During this process, two main challenges occur. Firstly, since the number of possible traffic situations is unlimited, a selection of a manageable number of representative situations to be tested must be applied during the certification of automated vehicles. Secondly, nowadays a limited number of standardized test cases are used for the type approval of vehicles. This can lead to so-called gaming of tests, which means that the manufacturer optimizes the system’s performance in the predefined test cases. A prominent example are the current discussions about the large differences between the emissions of vehicles in the driving cycle (e.g., WLTP) and in everyday use in road traffic. This paper addresses both stated challenges and exemplifies a method for the system-specific selection of test cases for the certification of automated vehicles, which are not known to the manufacturer in advance. Based on a system analysis and an objective driving behavior characterization, weak spots of the system under test are identified and connected to complex scenarios to be tested. This approach allows an economic and meaningful certification process for automated vehicles.