An innovative computational approach is proposed to design ionic liquids (ILs) based on a new a priori molecular descriptor of ILs, derived from quantum-chemical COSMO-RS methodology. In this work, the charge distribution on the polarity scale given by COSMO-RS is used to characterize the chemical nature of both the cations and anions of the IL structures, using simple molecular models in the calculations. As a result, a novel a priori quantum-chemical parameter, Sσ-profile, is defined for 45 imidazolium-based ILs, as a quantitative numerical indicator of their electronic structures and molecular sizes. Subsequently, neural networks (NNs) are successfully applied to establish a relationship between the electronic information given by the Sσ-profile molecular descriptor and the density properties of IL solvents. As a consequence, we develop here an a priori computational tool for screening ILs with required properties, using COSMO-RS predictions to NN design and optimization. Current methodology is validated following a classical quantitative structure−property relationship scheme, which is the main aim of this work. However, a second part of the current investigation will be devoted to a more useful design strategy, which introduces the desired IL properties as input into inverse NN, resulting in selections of counterions as output, i.e., directly designing ILs on the computer.