Abstract Background We investigate whether correct infant feeding knowledge and practice differ by maternal HIV status in an era of evolving clinical guidelines in rural South Africa. Methods This cohort study was nested within the MONARCH stepped-wedge cluster-randomised controlled trial (www.clinicaltrials.gov: NCT02626351) which tested the impact of continuous quality improvement on antenatal care quality at seven primary care clinics in KwaZulu-Natal, from July 2015 to January 2017. Women aged ≥18 years at delivery were followed up to 6 weeks postpartum. Clinical data were sourced from routine medical records at delivery. Structured interviews at early postnatal visits and the 6-week postnatal immunisation visit provided data on infant feeding knowledge and feeding practices respectively. We measured the relationship between maternal HIV status and (i) correct infant feeding knowledge at the early postnatal visit; and (ii) infant feeding practice at 6 weeks, using Poisson and multinomial regression models, respectively. Results We analysed data from 1693 women with early postnatal and 471 with 6-week postnatal interviews. HIV prevalence was 47% (95% confidence interval [CI] 42, 52%). Women living with HIV were more knowledgeable than women not living with HIV on correct infant feeding recommendations (adjusted risk ratio, aRR, 1.08, p < 0.001). More women living with HIV (33%; 95% CI 26, 41%) were not breastfeeding than women not living with HIV (15%; 95% CI 11, 21%). However, among women who were currently breastfeeding their infants, fewer women living with HIV (5%; 95% CI 2, 9%) mixed fed their babies than women not living with HIV (21%; 95% CI 14, 32%). In adjusted analyses, women living with HIV were more likely to avoid breastfeeding (adjusted relative risk ratio, aRRR, 2.78, p < 0.001) and less likely to mixed feed (aRRR 0.22, p < 0.001) than women not living with HIV. Conclusions Many mothers in rural South Africa still do not practice exclusive breastfeeding. Women living with HIV were more knowledgeable but had lower overall uptake of breastfeeding, compared with women not living with HIV. Women living with HIV were also more likely to practice exclusive breastfeeding over mixed feeding if currently breastfeeding. Improved approaches are needed to increase awareness of correct infant feeding and exclusive breastfeeding uptake.