This study aimed to investigate the association between consumption of ultra-processed foods, whole foods and breast cancer risk in black women from Soweto, South Africa. A population-based case (n 396)–control (n 396) study matched on age and residence, using data from the South African Breast Cancer study. Dietary intake was assessed using a validated quantified FFQ. Food items were categorised using the NOVA system ((1) unprocessed/minimally processed foods, (2) culinary ingredients, (3) processed foods and (4) ultra-processed foods). Conditional logistic regression models were used to estimate OR and 95 % CI of dietary contributions from each NOVA food group (as a percentage of total energy intake (EI)) and adjusting for potential confounders. Considering contributions to total EI per day, ultra-processed food consumption contributed to 44·8 % in cases and 47·9 % in controls, while unprocessed/minimally processed foods contributed to 38·8 % in cases and 35·2 % in controls. Unprocessed/minimally processed food consumption showed an inverse association with breast cancer risk overall (OR = 0·52, 95 % CI 0·35, 0·78), as well as in pre- and postmenopausal women separately (OR = 0·52, 95 % CI 0·27, 0·95 and OR = 0·55, 95 % CI 0·35, 0·89, respectively) and in women with progesterone positive breast cancer (OR = 0·23, 95 % CI 0·06, 0·86). There was no heterogeneity in association with breast cancer when analyses were stratified according to BMI. No significant associations were observed for the consumption of other NOVA food groups. Intake of unprocessed/minimally processed foods may reduce the risk of developing breast cancer in black women from Soweto, South Africa.