A high-strength concrete subjected to compressive fatigue loading with two maximum stress levels was investigated and the behaviour was evaluated using the macroscopic damage indicators, strain and acoustic emission hits (AE-hits), combined with microstructural analyses utilising light microscopy and scanning electron microscopy (SEM). A clustering technique using Gaussian mixture modelling combined with a posterior probability of 0.80 was firstly applied to the AE-hits caused by compressive fatigue loading, leading to two clusters depending on the maximum stress level. Only a few cracks were visible in the microstructure using light microscopy and SEM, even in phase III of the strain development, which is shortly before failure. However, bluish impregnated areas in the mortar matrix of higher porosity or defects, changing due to the fatigue loading, were analysed. Indications were found that the fatigue damage process is continuously ongoing on a micro- or sub-microscale throughout the mortar matrix, which is difficult to observe on a mesoscale by imaging. Furthermore, the results indicate that two different damage mechanisms take place, which are pronounced depending on the maximum stress level. This might be due to diffuse and widespread compressive damage and localised tensile damage, as the findings documented in the literature suggest.