Understanding the behaviour of malaria vectors could lead to improved designs of partially treated insecticidal nets and low-cost nets with less insecticide content. The behaviour of Anopheles arabiensis around cow baited bed nets with partial adhesive treatments were assessed in experimental huts. The study was conducted in Moshi Tanzania, using a Latin Square design with five arms: no bed-net, intact untreated bed-net, roof, sides and whole adhesive treated bed-nets. The data analyses were done using generalized linear mixed effects models with proportions of mosquitoes on the bed-net panels, and induced exiting as outcomes and trial arm as the predictor. There were significant reductions in the likelihood of An. arabiensis found on the side (adjusted odds ratio, AOR: 0.50; 95% CI 0.26-0.98; p value = 0.044) and roof (AOR: 0.18; 0.07-0.43; p value < 0.001) of adhesive treated bed-nets compared to the whole adhesive treated bed-net. The likelihood of An. arabiensis exiting was significantly higher in the intact untreated net trial arm compared to the no net trial arm (AOR: 2.12; 1.19-3.79; p value = 0.010). Host seeking An. arabiensis is not persistent and untreated bed-nets induced exiting, implying reduced efficacy of partially treated insecticidal nets.