A microwave-assisted extraction (MAE) technique was employed on grape skin pomaces to enable the extraction of different groups of phenolic compounds (total phenolics, tannins, flavonols, and hydroxycinnamic acids) and to obtain extracts with the highest antioxidant capacity (ORAC). The single-step extraction process was modeled and optimized by means of artificial neural network (ANN) and response surface methodology (RSM) coupled with full factorial design. Methanol concentration (20 to 100%, v/v), temperature (30 to 60°C) and duration (2 to 16 min) were MAE input parameters studied. Optimal parameters were further applied in multistep MAE cycles for the complete recovery of phenolic antioxidants. Results showed that methanol concentration was the most significant parameter influencing the extraction of all groups of phenolics and antioxidant capacity of extracts. Moreover, a significant effect of time and temperature was also noticed, except in the case of total hydroxycinnamic acids. The presented ANN model accurately predicted the effect of the three input parameters simultaneously on the output parameters (training R2=0.9957; test R2=0.9945; validation R2=0.9965). Optimal parameters showed that higher methanol concentrations and lower temperatures (100%, v/v; at 40°C) were more convenient for the extraction of flavonols and hydroxycinnamic acids than for ORAC (78.1%, v/v; at 60°C) or total phenolics and tannins (62.7 and 65.3%, v/v; at 60°C). The number of MAE cycles was found to be a key factor for completing extraction of skin pomace phenolics and should always be considered prior to analytical determination.