Achieving high coverage of antiretroviral treatment (ART) in resource-poor settings will become increasingly difficult unless HIV incidence can be reduced substantially. Universal voluntary counselling and testing followed by immediate initiation of ART for all those diagnosed HIV-positive (universal testing and treatment, UTT) has the potential to reduce HIV incidence dramatically but would be very challenging and costly to deliver in the short term. Early modelling work in this field has been criticised for making unduly optimistic assumptions about the uptake and coverage of interventions. In future work, it is important that model parameters are realistic and based where possible on empirical data. Rigorous research evidence is needed before the UTT approach could be considered for wide-scale implementation. This paper reviews the main areas that need to be explored. We consider in turn research questions related to the provision of services for universal testing, services for immediate treatment of HIV-positives and the population-level impact of UTT, and the research methods that could be used to address these questions. Ideally, initial feasibility studies should be carried out to investigate the acceptability, feasibility and uptake of UTT services. If these studies produce promising results, there would be a strong case for a cluster-randomised trial to measure the impact of a UTT intervention on HIV incidence, and we consider the main design features of such a trial.