BACKGROUND: We evaluated continuous quality improvement (CQI) targeting antenatal HIV care quality in rural South Africa using a stepped-wedge cluster-randomised controlled trial (Management and Optimisation of Nutrition, Antenatal, Reproductive, Child health, MONARCH) and an embedded process evaluation. Here, we present results of the process evaluation examining determinants of CQI practice and 'normalisation.' METHODS: A team of CQI mentors supported public-sector health workers in seven primary care clinics to (1) identify root causes of poor HIV viral load (VL) monitoring among pregnant women living with HIV and repeat HIV testing among pregnant women not living with HIV, and (2) design and iteratively test their own solutions. We used a mixed methods evaluation with from CQI mentors ('dose' and 'reach' of CQI, causes of poor HIV care testing rates, implemented change ideas); (HIV care testing by clinic and time step); and with available health workers. We analysed field notes andsemi-structured for determinants of CQI implementation and 'normalisation' using Normalisation Process Theory (NPT) and Tailored Implementation of Chronic Diseases (TICD) frameworks. RESULTS: All interviewed health workers found the CQI mentors and methodology helpful for quality improvement. Total administered 'dose' was higher than planned but 'reach' was limited by resource constraints, particularly staffing shortages. Simple workable improvements to identified root causes were implemented, such as a patient tracking notebook and results filing system. VL monitoring improved over time, but not repeat HIV testing. Besides resource constraints, gaps in knowledge of guidelines, lack of leadership, poor clinical documentation, and data quality gaps reduced CQI implementation fidelity and normalisation. CONCLUSION: While CQI holds promise, we identified several health system challenges. Priorities for policy makers include improving staffing and strategies to improve clinical documentation. Additional support with implementing clinical guidelines and improving routine data quality are needed. Normalising CQI may be challenging without concurrent health system improvements.