Excessive phosphorus (P) concentrations can lead to conditions that limit the amenity of freshwater resources. This problem is particularly acute in agricultural catchments, where P fertilizer and manure amendments have been used to increase soil fertility and productivity. In these catchments, P indices are often used to help target critical source areas in order to reduce P exports. However, the overall impact of agricultural mitigation efforts on receiving waters has not always been consistent with declines in total P exports from catchments. In this paper we propose a model of dissolved P mobilization (i.e., entrainment) in surface runoff that accounts for this outcome and examine modifications to P indices that better accommodate dissolved P mobilization. We suggest that dissolved P mobilization commences near the soil surface and has two phases. When water is first applied, labile P is mostly mobilized by dissolution and advection. Subsequently, as the supply of readily accessible P is exhausted, diffusion and hydrodynamic dispersion mobilize P from other sources at a near constant rate for the remainder of the event. As most P exports occur in larger (i.e., longer) events, the second phase appears responsible for most dissolved P exports. Such a model of dissolved P mobilization is consistent with runoff monitoring data under natural and simulated rainfall, suggesting that on low (shallow) slopes where the interaction between surface soil and water may be prolonged, dissolved P concentrations are likely to be higher. Dissolved P mobilization from low‐slope areas is not well represented in P indices at present. We suggest that there needs to be a more complex, mechanistic structure to P indices that involves additional compartmentalization. Further, we suggest that this can be achieved without losing the simplicity of P indices or flexibility to integrate research data and experiential knowledge into tools that are relevant to specific regions.