AbstractManagement of phosphorus (P) inputs to agroecosystems is often evaluated by phosphorus use efficiency (PUE), but different calculation approaches may yield discrepancies in estimated PUE values. We evaluated PUE (i) indirectly calculated by difference in crop P uptake between a P‐fertilized versus a P‐unfertilized control (PUEdiff) and (ii) directly measured by tracing radioisotopically labeled fertilizer P (PUEisotope). Using two ammonium phosphate fertilizers of high (monoammonium phosphate [MAP]) and low (struvite) water solubility in three soils with non‐limiting extractable soil P concentrations (Mehlich‐3 P > 25 mg kg−1), we find dissimilar PUEisotope versus PUEdiff. PUEdiff often yielded negative values (77% of observations), whereas PUEisotope was (i) positive and (ii) higher (two‐ to fourfold) for MAP than struvite, and (iii) unassociated with soil Mehlich‐3 P concentrations. Thus, PUEdiff appears to underestimate PUE and miss differences in PUE among fertilizerss. Our results raise the need to reconsider calculation of PUE and to address practical challenges to in situ measurement of PUEisotope.