Banded iron formations (BIFs) are marine chemical sedimentary rocks that were widely deposited before the start of the Great Oxidation Event (2.45–2.32 Ga). They represent an important archive into the chemistry of the Precambrian oceans and atmosphere, and the evolution of the marine biosphere. It is believed that BIFs were derived from ferric oxides/hydroxides that were deposited by iron-oxidizing bacteria or by reaction with oxygen released by Cyanobacteria. Hematite is traditionally interpreted to represent the dehydrated product of the ferric oxide/hydroxide precipitate, whereas magnetite is thought to have formed from ferric oxides/hydroxides via bacterial iron reduction and oxidation of microbial biomass. However, studies are equivocal about when and how magnetite formed in BIFs. We present the results of a detailed petrographic study of magnetite in BIFs of the Hamersley Group, Western Australia. We find no evidence that magnetite was precipitated from seawater or was a product of bacterial iron reduction. Instead, most of the magnetite formed by replacement of siderite and other ferrous-rich minerals after burial. Our observations suggest that magnetite growth was linked to the breakdown of siderite during metamorphism. Published experimental data indicate that in the presence of water and under the metamorphic conditions experienced in the southern Pilbara Craton (200–350 °C), the thermal decomposition of siderite could account for much of the magnetite in the Hamersley BIFs (3FeCO3 + H2O → Fe3O4 + 3CO2 + H2). The relationship between magnetite and the primary precipitates, which are preserved as greenalite nanoparticles and microgranules in chert bands, suggests that much of the magnetite formed in siderite-rich, iron-silicate bands that developed by compaction of non-silicified iron-silicate muds. These bands acted as reactive pathways for fluid flow and the formation of magnetite. If most of the magnetite formed after burial then it is necessary to reassess assumptions about the chemistry of the original precipitates and the role of microbial life in the redox cycling of iron. Our results are consistent with models for the deposition of BIFs that center around a precursor sediment comprising greenalite precipitates and diagenetic siderite that was partially replaced by magnetite and hematite after burial.