An artificial neural network (ANN) was used to predict the biosorption of methylene blue on Spirulina sp. biomass. Genetic and anneal algorithms were tested with different quantity of neurons at the hidden layers to determine the optimal neurons in the ANN architecture. In addition, sensitivity analyses were conducted with the optimised ANN architecture for establishing which input variables (temperature, pH, and biomass dose) significantly affect the predicted data (removal efficiency or biosorption capacity). A number of isotherm models were also compared with the optimised ANN architecture. The removal efficiency or the biosorption capacity of MB on Spirulina sp. biomass was adequately predicted with the optimised ANN architecture by using the genetic algorithm with three input neurons, and 20 neurons in each one of the two hidden layers. Sensitivity analyses demonstrated that initial pH and biomass dose show a strong influence on the predicted removal efficiency or biosorption capacity, respectively. When supplying two variables to the genetic algorithm, initial pH and biomass dose improved the prediction of the output neuron (biosorption capacity or removal efficiency). The optimised ANN architecture predicted the equilibrium data 5,000 times better than the best isotherm model. These results demonstrate that ANN can be an effective way of predicting the experimental biosorption data of MB on Spirulina sp. biomass.