The noise level in the gas microsensors is a tool for characterizing the electrical conduction under various gases and a means to improve selectivity. Metal-oxide gas microsensors with WO 3 sensitive thin film are characterized using a low-frequency noise technique. The spectral form of the noise responses measured using our specific systems is similar for tested gases (ozone and nitrogen dioxide). We observe a clear Lorentzian behavior according to adsorption-desorption (A-D) noise theory. To identify the detected gas, a physical-based characterization model of A-D noise source is proposed and compared with the empirical flicker noise model. We show that the excess noise is due to the A-D processes on the surface of the sensors sensitive film. The Lorentzian parameters depend on the nature of the gases and the noise-level dependence with gas concentration is clearly demonstrated. This confirms the interest on noise spectroscopy to improve the selectivity of gas sensors