Experiments on temperature polarization reduction in air gap membrane distillation were carried out. Dependence of permeate flux and pressure drop on the flow rate for five spacers of the same thickness with different geometries was investigated. The maximum flux with spacers was about 2.5 times higher, compared to an empty channel. Three different cases were considered: (1) when flux is of main importance, (2) when minimum pressure drop is required and flux is of less importance and (3) when a high flux-to-pressure-drop ratio is needed. Empty channel was found to be the best option for the second and third case, while being unacceptable in the first case. The spacer, made of round rods, having angle between the rods 90 degrees was found to be optimal for the first case, together with the spacer made of twisted tapes at angle of 120 degrees with flow attack angle of 30 degrees. It was established that four spacers out of five produced the same permeate flux for a given pressure drop value, in spite of the different dependencies, obtained for the flux and pressure drop. The correspondence between flux and pressure drop values for empty channel was found to be the same as for those four spacers in the range of low fluxes. This finding allows to assume that in the case of spacer-filled channels with not too high resistance analogy between momentum and heat transfer holds. It is concluded that at low fluxes the optimal spacer for air gap membrane distillation has to have the same dependence of flux on the pressure drop as an empty channel, but should be capable of producing sufficiently high fluxes at much lower flow rates than empty channel.