In industrial practice, it is common to use commercially available catalysts to facilitate chemical reactions. Unfortunately, due to the confidential composition and complex properties of the catalysts, a precise understanding of the reaction mechanism and intermediates is often difficult to obtain. This complicates the development of robust kinetic models, delaying an effective reactor design. This raises the question of whether machine learning (ML)-based regressions can reliably describe kinetics without requiring detailed reaction mechanisms. In this work, this research question was addressed by performing kinetic experiments on a commercial Ni/ZrO 2-based CO 2 methanation catalyst (Himetz by Kanadevia Corporation) in an isothermal, fixed bed reactor. A 216-point data set was thus obtained at various temperatures (220−300 °C), partial pressures (0.8−3 bar) and gas hourly space velocities (50,000−700,000 h −1) combinations. The data set was employed to perform several ML-based regressions, and to compare them in terms of adequacy of the fit, according to root mean squared error (RMSE). The two best performing ML-based models, Gaussian Process Regression and 1-layer neural network were selected and assessed against two standard kinetic modeling approaches: power law and Langmuir−Hinshelwood-Hougen-Watson (LHHW). The ML-based models outperformed the power law according to RMSE and were comparable to the LHHW model to describe the kinetic regime of the catalyst. When implemented in a reactor model, the ML-based regressions also accurately predicted methane yield with results comparable to the state-of-the-art LHHW model and even outperformed the LHHW model by addressing nonideal behavior in a nonisothermal reactor. This showed that ML-based kinetics are promising in situations where little mechanistic information is available, generating models that can be employed for reactor design purposes.