A deep understanding of the relationship between the properties and structure of transition metal oxide semiconductor gas sensors is essential for optimizing sensor design. Comparing and analyzing the sensing properties of nanomaterials with different morphologies and crystal facets by combining the experiment analysis and the density functional theory (DFT) calculations could identify and understand the corresponding mechanisms at the atomic or ionic level. Here, grown Co3O4 nanosheet (Co3O4-NS) with exposed (111) facets, and Co3O4 nanorods (Co3O4-NR) with exposed (110) facets on alumina ceramic tubes were prepared and studied in sensing ethanol. The adsorption characteristics of ethanol on the sensor surface were analyzed by DFT calculation. The results demonstrated that the sensing performance of Co3O4-NR was superior to Co3O4-NS. The adsorption of hydroxyl groups in ethanol was better than the methyl groups on the surface. The oxygen and hydrogen atoms from the hydroxyl groups in ethanol had the lowest adsorption energies (Eads = −3.44 eV) when they interact with the Co atom and the indirectly adjacent low-coordinate oxygen atoms on the (110) surface, respectively. The synergistic adsorption of target gases by exposed surface atoms induced superior sensing properties of the material.