The durability of fibre-reinforced polymer (FRP)-strengthened structural elements has been subject of recent research. Despite a large number of experimental studies in this field, there is no comprehensive model to predict the performance of composite systems exposed to severe environmental conditions. This paper evaluates the durability of FRP-to-concrete joints by applying machine learning (ML)-based data-driven techniques. A comprehensive experimental database on the durability of FRP-bonded connections subjected to water immersion conditions is collected to develop ML models. Three approaches, namely multiple linear regression (MLR), artificial neural networks (ANN) and adaptive-neuro fuzzy inference system (ANFIS) are established to identify the shear bond strength based on the geometrical and mechanical properties and the environmental conditions. The performance of these methods is evaluated based on their accuracy in predicting the bond strength of FRP-to-concrete joints after exposure to water immersion conditions. A comparison of the results indicates that ANN outperforms other methods with reasonable accuracy in estimating bond strength. Moreover, the study on the effects of environmental factors demonstrates that the combination of moisture and elevated temperature highly impacts the bond capacity of the joints immersed in water. Finally, an equation is proposed as a function of geometrical, mechanical and environmental properties of the FRP-to-concrete connections which can be used for rapid assessment of the bond capacity in adhesively bonded joints.