A stochastic approach is proposed to predict responses to selection when using αs1-casein genotype information in a selection scheme of a Spanish breed of dairy goats. Two independent selection objectives were considered: protein yield (PY), where the major additive gene CSNIS1, which codes for αs1-casein, has a small effect, and protein content (P%), where this gene has a large effect on performances. Significant differences in response between using and ignoring information on the major gene were observed only when the major gene has a large effect. The main result was in the case of P%, the total genetic gain obtained in the early generations of selection was maintained in the long-term. Taking account of genotype information either in the evaluation model or in the selection criteria leads to a faster fixation of the favourable allele and a reduction of the total genetic variance over generations. The inbreeding rates varied across generations, the highest rates observed in later generations of selection and when the major gene has a large effect and its genotype was included in the genetic evaluation procedure. It is concluded that inclusion of the casein genotype as an additional selection criteria will improve gains for protein traits, in particular P%. Recommendations are also given in order to optimize the use of this molecular information in dairy goat selection programs.