Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) investigations of molecules at the surface of catalysts exhibit a strong overlap of the adsorption peaks. Therefore, the investigation of the CO2 hydrogenation on a highly active catalyst surface requires a deconvolution of the adsorption spectra to clearly assign the signal to the chemical species. We developed an autonomous and efficient bi-level evolutionary Gaussian fitting (BEGF) procedure with a genetic algorithm at the upper level and a multipeak Gaussian fitting algorithm at the lower level to analyze self-consistently the set of spectra of an entire experiment. We show two examples of the application of BEGF procedure by analyzing the DRIFTS spectral sets of ex situ HCOO–* and CO2 hydrogenation on Ru/Al2O3. The fitting procedure deconvoluted the overlapped peaks and identified the bond vibrations of carbon monoxide, formate, bicarbonate, and carbonate through the developing trends of the peak intensities along the reaction. These revealed the progression of those species over the reaction timeline.