The gas sensor based on pristine graphene with conductance type was studied theoretically and experimentally. The time response of conductance measurements showed a quickly and largely increased conductivity when the sensor was exposed to ammonia gas produced by a bubble system of ammonia water. However, the desorption process in vacuum took more than 1 h which indicated that there was a larger number of transferred carriers and a strong adsorption force between ammonia and graphene. The desorption time could be greatly shortened down to about 2 min by adding the flow of water-vapor-enriched air at the beginning of the recovery stage which had been confirmed as a rapid and high-efficiency desorption process. Moreover, the optimum geometries, adsorption energies, and the charge transfer number of the composite systems were studied with first-principle calculations. However, the theoretical results showed that the adsorption energy between NH3 and graphene was too small to fit for the experimental phenomenon, and there were few charges transferred between graphene and NH3 molecules, which was completely different from the experiment measurement. The adsorption energy between NH4 and graphene increased stage by stage which showed NH4 was a strong donor. The calculation suggested that H2O molecule could help a quick desorption of NH4 from graphene by converting NH4 to NH3 or (NH3)n(H2O)m groups, which was consistent with the experimental results. This study demonstrates that the ammonia gas produced by a bubble system of ammonia water is mainly ammonium groups of NH3 and NH4, and the NH4 moleculars are ideal candidates for the molecular doping of graphene while the interaction between graphene and the NH3 moleculars is weak.