Detection and sensing of toxic gases are vital in many areas such as in environmental and safety monitoring, industrial applications and etc. Fundamental requirements for modern gas sensors typically are small size, low power consumption, high sensitivity and good selectivity towards various gases. Recently, graphene-based gas sensors have attracted significant attention due to their potential applications. Based on published works, the energy band gap of monolayer graphene nanoribbon (GNR) varies by changing its ribbon width. However, in the case of bilayer GNR, the energy band gap varies via three ways: the ribbon width, the layers separation and the symmetry between the layers. Therefore, it is interesting to investigate the characteristics of bilayer armchair graphene nanoribbon (B-AGNR) as gas nanosensor in presence of CO, O-2 and CO2 gas molecules. In this study, we report a theoretical investigation and numerical simulation of the electronic properties of monolayer and bilayer AGNR (with 10 atoms in width) and study their sensitivities in presence of CO, O-2 and CO2 gas molecules. Using Quantum espresso-5.3.0 package, the structure optimizations are done based on density functional theory (DFT). Moreover, we have evaluated the current-voltage (I-V) characteristics of the aforementioned structures using the open source package TRAN SIESTA which is based on non-equilibrium Green's function (NEGF) method. The obtained results show that the adsorption of CO, O-2 and CO2 gas molecules onto AGNR can be modified significantly in bilayer AGNR with respect to the monolayer. These gas molecules show weak interaction with the AGNR atomic structure due to large interaction distances. Thus, it is deduced that the graphene nanoribbon (AGNR) shows weak gas sensing in presence of CO, O-2 and CO2 molecules. Meanwhile, bilayer AGNR shows different electronic properties compared to monolayer AGNR which makes it highly sensitive and selective to the presence of O-2 and CO gases.