Objectives: To determine the upper limit for the incidence of clinically important HIV superinfection among HIV-infected men who have sex with men (MSM) and its relationship with engagement in unsafe sexual practices. Methods: This was a retrospective cohort and nested case-control study. Electronic files of all HIV-infected MSM not on antiretroviral therapy were reviewed. Those clients with sudden, unexplained and sustained declines in CD4 T-cell counts and increases in plasma HIV RNA were considered as being putatively superinfected with HIV and were recruited as cases, whereas those without these features were recruited as controls (four per case) to answer a self-administered questionnaire. Results: Ten cases were identified from 145 eligible MSM (7%, 95% confidence interval 3–11%), comprising a rate of 3.6 per 100 person-years at risk. Cases had an annual decline in CD4 T-cell counts of 201 cells µL–1 compared with 9 cells µL–1 for controls. There were no statistically significant differences between cases and controls with regard to sexual practices that may have exposed them to acquisition of HIV superinfection (P-value ≥ 0.4), nor in their perceptions or beliefs of HIV superinfection (P-value ≥ 0.3). Only a minority reported no previous knowledge of HIV superinfection (17%, 5/30). Overall, both cases and controls were engaging frequently in unsafe sexual practices with casual partners who were HIV infected (80 and 52%, respectively; P-value = 0.4) or whose HIV serostatus was unknown (40 and 50%, respectively; P-value = 1.0). Conclusions: Despite considerable unsafe sexual practices occurring among this cohort of sexually active MSM the incidence of clinically significant HIV superinfection was likely to be less than 4% per year.