The biological carbon pump (BCP) transfers CO₂ from the surface ocean to deep waters and has historically regulated atmospheric CO₂ levels, accounting for around one-third of glacial-interglacial CO₂ fluctuations. Current satellite-based methods for estimating BCP strength rely on empirical relationships between net primary production (NPP) and carbon export ratio. However, these methods contain a critical flaw: they cannot distinguish between organic matter that is quickly recycled near the surface and organic matter that reaches the deep ocean. Using an inverse biogeochemical model, we demonstrate that ~60% of satellite-measured NPP is rapidly recycled in the euphotic zone and never influences deep ocean tracer distributions. This finding has two important implications. First, the true strength of the BCP is better constrained by deep ocean tracer distributions than by satellite measurements - our model shows consistent export rates regardless of which satellite NPP product is used. Second, while deep tracers robustly constrain the mean state of the BCP, they are less useful for capturing temporal variability. We show that combining satellite observations with inverse modeling could help detect interannual variations in BCP strength, though longer observational records of both satellite and hydrographic data will be needed to make these temporal signals statistically significant.