Context. Deriving the metallicity, [Fe/H], in low-resolution spectra of carbon-enhanced metal-poor (CEMP) stars is a tedious task that, owing to the large number of line blends, often leads to uncertainties on [Fe/H] exceeding 0.25 dex. The CEMP stars increase in number with decreasing [Fe/H] and some of these are known to be bona fide second generation halo stars. Hence, knowing their [Fe/H] is important for tracing the formation and chemical evolution of the Galaxy.Aims. Here, we aim to improve the [Fe/H] measurements in low-resolution spectra by avoiding issues related to blends. In turn, we improve our chemical tagging in such spectra at low metallicities.Methods. We developed an empirical way of deriving [Fe/H] in CEMP (and C-normal) stars that relates the equivalent width (EW) of strong lines, which remain detectable in lower resolution, metal-poor spectra, such as X-shooter spectra to [Fe/H].Results. The best [Fe/H] tracers are found to be Cr I and Ni I, which both show strong transitions in spectral regions that are free of molecular bands (between 5200-6800 angstrom, a region accessible to most surveys). We derive different relations for dwarfs and giants. The relations are valid in the ranges - 3 <[Fe/H] <-0.5 and 10 <EW <800 m angstrom (Cr) or [Fe/H] > -3.2 and EW > 5 m angstrom (Ni), depending on the trace element and line as well as the stellar evolutionary stage.Conclusions. The empirical relations are valid for both CEMP and C-normal stars and have been proven to be accurate tracers in a sample of 400 stars (mainly giants). The metallicities are accurate to within +/- 0.2 dex depending on the sample and resolution, and the empirical relations are robust to within 0.05-0.1 dex. Our relations will improve the metallicity determination in future surveys, which will encounter a large number of CEMP stars, and will greatly speed up the process of determining [Fe/H] as the EWs only need to be measured in two or three lines in relatively clean regions compared to dealing with numerous blended Fe lines.