In this work, a response surface methodology (RSM) and artificial neural network (ANN) were used to study the ultrasound-assisted simultaneous removal of ternary toxic dyes onto copper sulfide nanoparticles loaded on activated carbon (CuS–NP–AC), while dyes contentment analysis were undertaken by derivative spectrophotometry. The influence of process variables (adsorbent mass, sonication time, MG, DB and MB concentration dyes on adsorption was investigated by central composite rotatable design (CCRD) of RSM. The experimental data obtained through CCRD was used to train the ANN model. The results reveal satisfactory results of empirical models (p < 0.0001) for fitting to the experimental data. According to constructed the response model the maximum removal efficiency close to 100% for all dyes at following optimum value of operating variables: 10 mg/L of MG, DB and MB, 6 min sonication time: and 0.02 g CuS–NP–AC, at pH 8.0 was achieved. The ANN based on the Levenberg–Marquardt algorithm (LMA) composed of linear transfer function (purelin) at output layer and tangent sigmoid transfer function (tansig) at hidden layer with 5, 8 and 9 neurons for MG, DB and MB, respectively, give best operation conditions for good prediction of behavior adsorption system. The maximum adsorption capacity of CuS–NP–AC according to Langmuir isotherm model was 263.2, 243.9 and 204.1 mg/g for MG, DB and MB, respectively. The pseudo-second-order model extensively with high ability is able to predict behavior of dyes adsorption onto adsorbent.