This study proposes an animal selection protocol for adaptability using machine learning models to analyse variables related to genotype–environment interaction in cows raised in the Ñeembucú wetlands of Paraguay. The objective is to optimise selection and improve reproductive efficiency by addressing adaptive traits related to specific environments. Machine learning enabled the identification of key physiological variables associated with environmental adaptability that influence body condition in cows, including phosphatase, cholesterol, phosphorus, hair length, creatinine, haematocrit, creatine phosphokinase, haemoglobin, body temperature and calcium. The gradient boosting machine model was selected for its superior performance based on root mean square error and mean absolute error indicators. Results indicated that low concentrations of phosphatase and creatine phosphokinase, along with shorter hair length, positively affect body condition score. Likewise, body temperature dynamics were reflected in the response variable. Higher levels of haematocrit and haemoglobin showed a positive influence on body condition score. Based on the identified influential variables, a selection protocol for adaptability in breeding cows is proposed.