Gas transport in a coal matrix is critical for the evaluation of coalbed methane storage and gas extraction. This transport behavior is generally considered to be in accordance with Fick's law but may lead to some prediction errors because of the misconception that the adsorbed gas is also involved in the flow. In this work, a density gradient-driven model of gas adsorption within the coal matrix was developed under confined-space adsorption conditions, and a new parameter named the microchannel diffusion coefficient was proposed to characterize the transport capability of gas. This new model was compared with some common models, in which their simulated adsorption curves were matched with isothermal adsorption data to distinguish the optimal model. The results show that (i) compared to Fick's model, the simulated curves of the density gradient model are in better agreement with the experimental data throughout the adsorption process; (ii) the permeability of Darcy's model is heavily influenced by adsorption pressures, while the microchannel diffusion coefficients are more realistic and independent of time and pressure. Therefore, the density gradient model is better suited to describe the transport behavior of gas in microchannels, which helps to understand the essence of gas transport in coal matrices.