Mathematical modelling was combined with experimental Cu isotope measurements to demonstrate the effect of the sample matrix in changing the absolute and relative abundances of spectral interferences from Ti and Cr species. This unforeseen non-spectral effect, evidenced by variable inaccuracies of the different Zn-normalised Cu isotope ratios, was investigated by comparing real sedimentary samples and artificial solutions intended to match the Cu:Ti:Cr ratios of the real samples after (one or two step) chromatographic processing. Artificial solutions showed positive bias in δ65CuX/Y with the magnitude depending on (a) the 6XZn/6YZn ratio used for normalisation, (b) the Ti/Cu ratio and (c) the transmission coefficient of the TiO species. In contrast, real samples showed different δ65CuX/Y patterns and displayed a more complex population of Ti and Cr oxides and hydroxides, giving rise to positive and negative inaccuracies that were two to six times higher compared with the artificial samples. The results evidence contrasting behaviour of Ti and Cr when forming polyatomic species in the plasma and stress that artificial solutions may fail to predict how residual elements interact with the analyte/dopant pair during MC-ICP-MS analyses. More importantly, the study shows that all Zn isotope ratios do not have the same merit in correcting for mass bias in the presence of matrix elements and should all be monitored to verify the absence of spectral interferences for Cu isotope measurements. In this respect, accurate Cu data could be generally obtained by a two-step chromatographic purification providing a minimum reduction of ~ 21000 and ~ 3000 times the initial amounts of Ti and Cr, respectively.