In this study, effects of zeta potential distribution and geometrical specifications are numerically investigated on mixing efficiency in electroosmotic flows. Considered geometries include straight, converging, diverging, and converging-diverging microchannels. Electroosmotic flow simulations are conducted based on the N-S and Nernst-Planck equations for momentum and ionic charges distributions, respectively, by lattice Boltzmann method. Numerical simulations are validated against available analytic electroosmotic flow solutions in homogeneous straight channels, and then flow patterns and mixing performances in the presence of non-uniform zeta potential distributions are investigated in search for enhanced mixing performances. Numerical results indicate that converging channel leads to a sizable increase in mixing efficiencies, while the flow rate decreases at the same time. In contrast, diverging channels increase the flow rate, while decrease the mixing efficiency. Therefore, it is expected to achieve a balance between the mixing efficiency and mass flow rate using converging-diverging geometries. Numerical results indicate that mixing efficiency of about 90% can be reached with a converging-diverging microchannel with a reasonable decrease in mass flow rate as compared to its geometrical diverging-converging counterpart channel.