The study of models for the correlation of the surface tension of fluids is a crucial issue in various fields of chemical industries. Usually, seeking for appropriate models is an arduous process that relies too much on the researcher's inventiveness. The main advantage of symbolic regression (SR) is being a straightforward method which automatizes the searching for models, and provides precise and trustworthy models for the researcher to work with. Although in this work we apply SR to develop correlation models for the surface tension of alcohols, the methodology is clearly applicable to any other thermodynamic property or any other family of fluids. We employ a database set of 87 alcohols, with a total quantity of 3570 data that have been carefully selected and filtered in previous works. We consider all the parameters which have a physical meaning for the model, and make a correlation study in order to select the most representative ones: temperature, critical temperature, critical pressure, critical volume, molar volume and acentric factor. Then we make the comparison for the best models obtained with SR (attending to accuracy and complexity of the model) and the usually employed polynomial regression models, obtaining lower errors (1−R2) with SR. Afterwards, we analyze and optimize the models offered by SR which include the least number of parameters possible and provide the highest accuracy. The best models offer values under 7.8% for MAPD (the lowest one is 6.8%), and under 0.07 for 1−R2 (the lowest one is 0.04).