Up-to-date, nationally representative household income/expenditure data are crucial to estimating poverty during the COVID-19 pandemic and to policy-making more broadly, but South Africa lacks such data. We present new pandemic poverty estimates, simulating incomes in pre-pandemic household surveys using contemporary labour market data to account for job losses between 2020 Q1 and 2021 Q4. Improving on much of the existing literature, we use observed rather than simulated shocks and allow for uneven impacts of the pandemic by employment sector and demographic characteristics. We present three updating methods, using the National Income Dynamics Study (NIDS) Wave 5, the Living Conditions Survey 2014/15, and the Quarterly Labour Force Survey (QLFS). Giving primacy to NIDS Wave 5 produces the largest estimate of pandemic period job-loss-induced poverty: a headcount ratio increase at the upper-bound poverty line of 5.2 percentage points (3.1 million people/13 per cent) and poverty gap increase of 3.8 percentage points (21 per cent). Giving primacy to QLFS data produces the lowest estimated change: a headcount ratio increase of 3.0 percentage points (1.8 million people/7 per cent) and poverty gap increase of 2.5 percentage points (12 per cent). Simulating receipt of the Special COVID-19 Social Relief of Distress social grant substantially mitigates poverty effects, with a poverty headcount increase of 1.1–3.4 percentage points and a poverty gap increase of 0.2–1.5 percentage points.