Surface complexation modelling (SCM) is a powerful tool to estimate speciation and fate of solutes in soil, provided sufficient model validation. This study aims to describe phosphate (PO4) leaching with SCM. The leachate phosphorus concentrations ([P]) of 120 unsaturated columns of contrasting agricultural soils were measured and modelled. Leachate [P] ranged 0.7–240 μM. Leachate [P] increased as the ratio of P to iron and aluminium ([Formula presented]) in acid oxalate soil extracts increased and as leachate Fe and Al concentrations ([Al + Fe]) increased. SCM was used to describe PO4 sorption to ferrihydrite (CD-MUSIC model). This yielded adequate description of leachate [P] (RMSElog10 = 0.39), but only when reactive PO4 was described from isotopically exchangeable PO4, when organic matter was included as the main competing adsorbate and when mobile colloidal ferrihydrite was included. The model reveals that colloidal PO4 transport enhanced leachate PO4 concentrations up to a factor 50 at small soil P content and small calcium (Ca2+) concentration in solution, as a large Ca2+ concentration enhances colloidal stability. This modelling approach explained that long-term application of organic fertilisers with higher Ca content reduced P leaching, likely due to the effect of Ca2+ on colloidal stability. A two-parameter empirical Langmuir model, based on soil Fe and Al oxyhydroxides, fitted data better than any SCM, suggesting that the empirical model might be advocated for application at large scale. This study revealed the power of SCM to better understand colloidal transport of P in soil.