The carbon-removal efficiency of ocean alkalinity enhancement (OAE) depends not only on alkalinity release, but also on its persistence in seawater. Yet mineral feedstocks are commonly evaluated under controlled dissolution conditions that may not represent alkalinity retention in dynamic coastal environments. Here, we quantified total alkalinity (TA) gains across complementary experimental systems using the same 95% pure Mg(OH)₂ feedstock in subtropical coastal seawater. We compared dark, well-mixed 55 L dissolution tests with naturally illuminated 55 L microcosms and 50,000 L open mesocosms under different seasonal conditions. In dissolution tests, the observed ΔTA matched the theoretical addition. Peak ΔTA represented 87–94% of the theoretical addition in microcosms, but only 58–59% in both mesocosm experiments despite contrasting seasonal conditions. Furthermore, after 19–35 days, control-referenced ΔTA represented only 32–51% of the theoretical addition which is insufficient to drive a net CO₂ sink. This sustained decline could not be attributed to a single process. Serpulid calcification was a biological sink, but inorganic CaCO₃ formation was also evident-despite experiments remaining below the critical aragonite saturation states for CaCO₃ formation in filtered seawater. A critical research challenge is distinguishing inefficient alkalinity dissolution from CaCO₃ formation. Reporting these quantities separately, together with constraints on mineral phases and biological sinks, is essential before controlled dissolution tests can be extrapolated to coastal OAE performance and carbon-removal efficiency.