ABSTRACT We present and discuss the results of a search for extremely metal-poor stars based on photometry from data release DR1.1 of the SkyMapper imaging survey of the southern sky. In particular, we outline our photometric selection procedures and describe the low-resolution (R ≈ 3000) spectroscopic follow-up observations that are used to provide estimates of effective temperature, surface gravity, and metallicity (Fe/H) for the candidates. The selection process is very efficient: of the 2618 candidates with low-resolution spectra that have photometric metallicity estimates less than or equal to −2.0, 41 per cent have Fe/H ≤ −2.75 and only approximately seven per cent have Fe/H > −2.0 dex. The most metal-poor candidate in the sample has Fe/H < −4.75 and is notably carbon rich. Except at the lowest metallicities (Fe/H < −4), the stars observed spectroscopically are dominated by a ‘carbon-normal’ population with C/Fe1D, LTE ≤ +1 dex. Consideration of the A(C)1D, LTE versus Fe/H1D, LTE diagram suggests that the current selection process is strongly biased against stars with A(C)1D, LTE > 7.3 (predominantly CEMP-s) while any bias against stars with A(C)1D, LTE < 7.3 and C/Fe1D,LTE > +1 (predominantly CEMP-no) is not readily quantifiable given the uncertainty in the SkyMapper v-band DR1.1 photometry. We find that the metallicity distribution function of the observed sample has a power-law slope of Δ(Log N)/ΔFe/H = 1.5 ± 0.1 dex per dex for −4.0 ≤ Fe/H ≤ −2.75, but appears to drop abruptly at Fe/H ≈ −4.2, in line with previous studies.