Circuit/model identification is applied to impedance spectroscopy when there is no prior knowledge of the inner workings of the sensor or process under analysis. This paper presents a performance assessment of gene expression programming for automatic circuit model identification in impedance spectroscopy. The main objective of this work is to improve gene expression programming specific implementation details, with pre-embedding knowledge regarding circuit simplification rules, in order to improve its performance for circuit identification in impedance spectroscopy. Three different versions of gene expression programming are presented, discussed and analyzed. Insight is given into the inner workings of gene expression programming while highlighting the proposed changes in the three versions. The performance of the improved algorithm is analyzed through numerical simulated impedance data. It is further validated by the successful application to measurements of a real sensor along with a study of its performance on measured data.