Most of the product patents claim a large number of compounds based on a Markush structure. However, the identification and optimization of new principal active ingredients is frequently driven by a simple Free Wilson approach, leading to a highly focused study only involving the chemical space nearby a hit compound. This fact raises the question: do the tested compounds described in patents really reflect the full molecular diversity described in the Markush structure? In this study, we contrast the performance of rational selection to conventional approaches in seven real-case patents, assessing their ability to describe the patent’s chemical space. Results demonstrate that the integration of computer-aided library selection methods in the early stages of the drug discovery process would boost the identification of new potential hits across the chemical space.