In order to maximize the effectiveness of “Seek, Test, and Treat” strategies for curbing the HIV epidemic, new approaches are needed to increase the uptake of HIV testing services, particularly among high-risk groups. Low HIV testing rates among such groups suggest that current testing services may not align well with the testing preferences of these populations. Female bar workers and male mountain porters have been identified as two important high-risk groups in the Kilimanjaro Region of Tanzania. We used conventional survey methods and a discrete choice experiment (DCE), a preference elicitation method increasingly applied by economists and policy-makers to inform health policy and services, to analyze trade-offs made by individuals and quantify preferences for HIV testing services. Bivariate descriptive statistics were used to analyze differences in survey responses across groups. Compared to 486 randomly selected community members, 162 female bar workers and 194 male Kilimanjaro porters reported 2–3 times as many lifetime sexual partners (p < 0.001), but similar numbers of lifetime HIV tests (median 1–2 across all groups). For the DCE, participants' stated choices across 12,978 hypothetical HIV testing scenarios (422 female and 299 male participants × 9 choice tasks × 2 alternatives) were analyzed using gender-specific mixed logit models. Direct assessments and the DCE data demonstrated that barworkers were less likely to prefer home testing and were more concerned about disclosure issues compared with their community counterparts. Male porters preferred testing in venues where antiretroviral therapy was readily available. Both high-risk groups were less averse to traveling longer distances to test compared to their community counterparts. These results expose systematic differences in HIV testing preferences across high-risk populations compared to their community peers. Tailoring testing options to the preferences of high-risk populations should be evaluated as a means of improving uptake of testing in these populations.