Although molybdenum (Mo) isotopic compositions of carbonatites, phosphorites and siliceous rocks can be used as proxies to reconstruct conditions of marine chemistry throughout geological time, only a few studies have, so far, analysed these low-Mo (ng g(-1)) geological samples, because of analytical limitations. In this study, a low blank, high yield two-column Mo purification procedure was developed for various low-Mo geological samples. The sample-standard bracketing (SSB) and double-spike (DS) methods for mass fractionation correction were used to compare the accuracy of Mo isotope ratio measurements. Six Mo reference materials, NIST SRM 3134 Mo, JMC Mo, SC+1 and SC-1 (eluted fractions of Sigma-Aldrich Mo), CRM GSR-6 limestone and USGS BCR-2 basalt, were used as quality controls. The results showed that the Mo delta values of reference materials and geological samples corrected by the SSB and DS methods were, within error, consistent with each other and the DS method was the method of choice for samples with <0.5 mu g g(-1) Mo. The average instrument long-term (over 1 year) external reproducibility of NIST SRM 3134 Mo was better than +/- 0.03 parts per thousand amu(-1) (2SD, n = 288) and the analytical precision of low-Mo (10(1) to 10(3) ng g(-1)) geological samples was better than +/- 0.04 parts per thousand amu(-1). This method can facilitate Mo isotope ratio measurements in geological samples with a low Mo content, offering a possibility to study a wider range of Mo reservoirs in geological processes.