In this work, phosphate removal was evaluated through a fixed-bed column. Experiments were carried out using three different adsorbents: raw dolomite (RD), calcined dolomite (CD), and ultrasound-modified dolomite (UD). An experimental design, Bayesian statistics, and data prediction were performed by modeling the breakthrough curves to study the efficiency of the process. To optimize the experimental process, the effects of initial phosphate concentration (15−60 mg L −1 ), adsorbent mass (0.5−3 g), and feed-flow rate (2.5−10 mL min −1 ) on saturation time and saturation concentration were studied. The results showed that the breakthrough curves' behavior for removing phosphate was similar for the solids tested. The Yoon−Nelson model represented better the mechanisms of the breakthrough curves for RD and UD, while for the CD, the Clark model had a better adjustment and the method could predict the breakthrough curves. Besides, UD was the most adequate adsorbent, resulting in better adsorption capacities. The best operational conditions for phosphate removal in UD were C i = 60 mg L −1 , W = 0.5 g, and Q = 10 mL min −1 , which resulted in 14.38 mg g −1 adsorption capacity and t sat = 25 min, with a superior value of q max of around 408.13% when compared with the lowest process time of the solid (2.83 mg g −1 with t sat = 17.5 min). Finally, the Gompertz model shows that intermediate values of α G and β G (around α G = 3.21 and β G = 0.27) increased the q max (14.38 mg g −1 ). It can be used for practical approaches, e.g., in the implementation of this process on an industrial scale.