Topical repellents provide protection against mosquito bites and their efficacy is often assessed using the arm-in-cage test. The arm-in-cage test estimates the repellent’s protection time by exposing a repellent-treated forearm to host-seeking mosquitoes inside a cage at regular intervals until the first confirmed mosquito bite. However, the test does not reveal the repellents’ behavioural mode of action. To understand how mosquitoes interact with topical repellents in the arm-in-cage test, we used a 3D infrared video camera system to track individual Aedes aegypti and Anopheles stephensi females during exposure to either a repellent-treated or an untreated forearm. The repellents tested were 20% (m/m) ethanolic solutions of N, N-diethyl-meta-toluamide, p-menthane-3,8-diol, icaridin and ethyl butylacetylaminopropionate. All four repellents substantially reduced the number of bites compared to an untreated forearm, while the flight trajectories indicate that the repellents do not prevent skin contact as the mosquitoes made multiple brief contacts with the treated forearm. We conclude that, in the context of the arm-in-cage test, topical repellents activate mosquitoes to disengage from the forearm with undirected displacements upon contact rather than being repelled at distance by volatile odorants.