The effect of inert sites in the global oscillations in the oxidation of CO on Pt(100) for both random and clustered inert sites is simulated by use of the cellular automaton technique. The cellular automaton rules account for the structural phase transformations of the Pt substrate, the reaction kinetics of the adsorbed phase and diffusion of adsorbed species. The introduction of a fraction θd of inert sites reduces the extent of the oscillatory region on the bifurcation diagram. The effect of added impurities on the transition from oscillatory to CO poisoned state is found to be quite different depending on the initial position on the bifurcation diagram. Points located near the rightmost branch on the bifurcation diagram will reach inert state through a very abrupt transition, whereas points located far from it undergo a much softer transition. Clustered impurities tend to soften the transition, and higher values of θd are needed to trigger the transition to the poisoned state. The study of a mean-field model without the oscillatory kinetics explains some of the basic features of the effect of increasing densities of impurities on the catalytic surface, for instance the dependence of the location of the transition point θd on the adsorption rate of CO.