Carbon dioxide (CO₂) conversion in non-thermal plasma is a promising route for carbon utilisation under mild conditions. This study investigates the performance and dynamic behaviour of kaolin-based catalysts modified with Ni (nickel), Ni–Ce (nickel-cerium), and Fe–Cu (iron-copper) oxides in a Dielectric Barrier Discharge (DBD) reactor. Materials were characterised using X-ray diffraction, energy-dispersive X-ray fluorescence, and scanning electron microscopy with energy-dispersive X-ray spectroscopy. CO₂ conversion was evaluated at varying Plasma Energy Numbers (PEN = 1.65–20) with time-resolved gas analysis over a 10 min period. Results demonstrate that the kaolin support is not inert; its dielectric properties actively influence discharge characteristics. Ni-based catalysts exhibited the highest stable activity, reaching ~53% conversion for samples calcined at 500 °C. Conversely, adding cerium oxide significantly decreased conversion and induced temporal instabilities, contrasting with its typical role in thermal catalysis. Time-resolved measurements revealed that Ni–Ce and Fe–Cu systems exhibit initial activity followed by gradual deactivation, suggesting plasma-induced surface restructuring. These findings highlight that catalyst performance in DBD is governed by a complex interplay of chemical activity and plasma–material interactions. The generated time-series data provide a robust foundation for machine learning applications in predictive modelling and stability classification of plasma-catalytic systems.