The paper describes the utilization of evolutionary algorithms for automatic discovery of benchmark circuits. The main objective of the paper is to show that relatively large and complex (benchmark) circuits can be evolved in case that only a given property (e.g. testability) is required and the function of the circuit is not considered. This principle is demonstrated on automatic discovery of benchmark circuits with predefined structural and diagnostic properties. Fitness evaluation for the proposed algorithm is based on testability analysis with linear time complexity. During the evolution, the solutions which are refused to be synthesized by a design system are excluded from the process of developing a new generation of benchmark circuits. The evolved circuits contain thousands of components and satisfy the required testability properties.