Reconstruction of the evolutionary history of Earth's habitability has been approached with various geochemical proxies of sedimentary rocks, particularly carbonates. It is a consensus that marine carbonates recorded the seawater geochemical compositions unless the primary signals were altered in diagenesis. This argument is the cornerstone for the applications of carbonate-based geochemical proxies on chemostratigraphic correlations and paleoenvironment interpretations. However, this assumption is only valid for carbonate precipitation in the water column, whilst most carbonate rocks outcropped in continents were precipitated on the seafloor. The modern seafloor commonly displays upward diffusions of various reduced components, e.g., Fe2+, Mn2+, H2S, and dissolved inorganic carbon (DIC), from sediment porewater to seawater, collectively called the benthic flux. As such, carbonates precipitated on the seafloor are inevitably affected by benthic flux from porewater, recording mixed signals of both seawater and porewater. In this study, we first reviewed the precipitation modes of marine carbonate. Then we explore how carbonate geochemical compositions could be affected by benthic flux when precipitated on the seafloor, and suggest using ferrous Fe content in carbonate (Fecarb) as a proxy to constrain benthic flux. Finally, with the consideration of benthic flux, we evaluate the traditional interpretation of carbonate carbon isotope and sulfur isotope of carbonate associated sulfate (CAS). We show that carbonate geochemical compositions do not reflect the seawater composition unless the benthic flux is negligible. Nevertheless, the seawater geochemical compositions, which could be used for regional or global chemostratigraphic correlations, could be retrieved by excluding benthic flux signals, whereas the extracted benthic flux signals could be used to confine the seafloor redox conditions. We suggest that interpretations of carbonate geochemical data should base on rigorous petrographic and sedimentological analyses.