In this study, the accuracy and the precision corresponding to Li isotopic measurements of low level samples such as marine and coastal carbonates are estimated. To this end, a total of fifty‐four analyses of a Li‐pure reference material (Li7‐N) at concentrations ranging from 1 to 6 ng ml−1 were first performed. The average δ7Li values obtained for solutions with and without chemical purification were 30.3 ± 0.4‰ (2s, n = 19) and 30.2 ± 0.4‰ (2s, n = 36), respectively. These results show that the chosen Li chemical extraction and purification procedure did not induce any significant isotope bias. Two available carbonate reference materials (JCt‐1 and JCp‐1) were analysed, yielding mean δ7Li values of 18.0 ± 0.27‰ (2s, n = 6) and 18.8 ± 1.8‰ (2s, n = 9), respectively. Small powder aliquots (< 15 mg) of JCp‐1 displayed significant isotope heterogeneity and we therefore advise favouring JCt‐1 for interlaboratory comparisons. The second part of this study concerns the determination of δ7Li value for biogenic carbonate samples. We performed a total of twenty‐nine analyses of seven different tropical coral species grown under controlled and similar conditions (24.0 ± 0.1 °C). Our sample treatment prior to Li extraction involved removal of organic matter before complete dissolution in diluted HCl. Our results show (a) a constant δ7Li within each skeleton and between the different species (δ7Li = 17.3 ± 0.7‰), and (b) a Li isotope fractionation of −2‰ compared with inorganic aragonite grown under similar conditions. Comparison with literature data suggests a significant difference between samples living in aquaria and those grown in natural conditions. Finally, we investigate ancient (fossil) carbonate material and foraminifera extracted from marine sedimentary records. Different leaching procedures were tested using various HCl molarities. Results indicate that carbonate preferential dissolution must be carried out at an acid molarity < 0.18 mol l−1. Possible contamination from silicate minerals can be verified using the Al/Ca ratio, but the threshold value strongly depends on the carbonate δ7Li value. When the silicate/carbonate ratio is high in the sediment sample (typically > 2), contamination from silicates cannot be avoided, even at low HCl molarity (⪡ 0.1 mol l−1). Finally, bulk carbonate and foraminifera extracted from the same core sample exhibited significant discrepancies: δ7Li values of foraminifera were more reproducible but were significantly lower. They were also associated with lower Sr/Ca and higher Mn/Ca ratios, suggesting a higher sensitivity to diagenesis, although specific vital effects cannot be fully ruled out.