BACKGROUND: Stigma, criminalisation and a lack of data on drug use contribute to the "invisibility" of people who inject drugs (PWID) and make HIV prevention and treatment service delivery challenging. We aimed to confirm locations where PWID congregate in Cape Town, eThekwini and Tshwane (South Africa) and to estimate PWID population sizes within selected electoral wards in these areas to inform South Africa's first multi-site HIV prevention project for PWID. METHODS: Field workers (including PWID peers) interviewed community informants to identify suspected injecting locations in selected electoral wards in each city and then visited these locations and interviewed PWID. Interviews were used to gather information about the accessibility of sterile injecting equipment, location coordinates and movement patterns. We used the Delphi method to obtain final population size estimates for the mapped wards based on estimates from wisdom of the crowd methods, the literature and programmatic data. RESULTS: Between January and April 2015, we mapped 45 wards. Tshwane teams interviewed 39 PWID in 12 wards, resulting in an estimated number of accessible PWID ranging from 568 to 1431. In eThekwini, teams interviewed 40 PWID in 15 wards with an estimated number of accessible PWID ranging from 184 to 350. The Cape Town team interviewed 61 PWID in 18 wards with an estimated number of accessible PWID ranging between 398 and 503. Sterile needles were only available at one location. Almost all needles were bought from pharmacies. Between 80 and 86% of PWID frequented more than one location per day. PWID who reported movement visited a median of three locations a day. CONCLUSIONS: Programmatic mapping led by PWID peers can be used effectively to identify and reach PWID and build relationships where access to HIV prevention commodities for PWID is limited. PWID reported limited access to sterile injecting equipment, highlighting an important HIV prevention need. Programmatic mapping data show that outreach programmes should be flexible and account for the mobile nature of PWID populations. The PWID population size estimates can be used to develop service delivery targets and as baseline measures.