BACKGROUND: Long-distance truck drivers in Africa are particularly at risk of HIV acquisition and offering self-testing could help increase testing coverage in this hard-to-reach population. The aims of this study are twofold: (1) to examine the preference structures of truck drivers in Kenya regarding HIV testing service delivery models and what they mean for the roll-out of HIV self-testing, and (2) to compare the preference data collected from a hypothetical discrete choice experiment with the actual choices made by participants in the intervention arm of a randomised controlled trial (RCT) who were offered HIV testing choices. METHODS: Using data from 150 truck drivers, this paper examines whether the stated preferences regarding HIV testing in a discrete choice experiment predict the actual test selected when offered HIV testing choices. Conditional logit models were used for main effects analysis and stratified models were run by HIV testing choices made in the trial to assess if the attributes preferred differed by test chosen. RESULTS: The strongest driver of stated preference among all participants was cost. However, two preferences diverged between those who actually chose self-testing in the RCT and those who chose a provider administered test: the type of test (p < 0.001) and the type of counselling (p = 0.003). Self-testers preferred oral-testing to finger-prick testing (OR 1.26 p = 0.005), while non-self-testers preferred finger-prick testing (OR 0.56 p < 0.001). Non-self-testers preferred in-person counselling to telephonic counselling (OR 0.64 p < 0.001), while self-testers were indifferent to type of counselling. Preferences in both groups regarding who administered the test were not significant. CONCLUSIONS: We found stated preference structures helped explain the actual choices participants made regarding the type of HIV testing they accepted. Offering oral testing may be an effective strategy for increasing willingness to test among certain groups of truck drivers. However, the importance of in-person counselling and support, and concern that an oral test cannot detect HIV infection may mean that continuing to offer finger-prick testing at roadside wellness centres will best align with the preferences of those already attending these facilities. More research is needed to explore whether who administers the HIV test (provider versus self) makes any difference. TRIAL REGISTRATION: This trial is registered with the Registry for International Development Impact Evaluations ( RIDE ID#55847d64a454f ).