The aim of this study was to examine how running biomechanics (spatiotemporal and kinetic variables) adapt with exhaustion during treadmill runs at 90, 100, 110, and 120% of the peak aerobic speed (PS) of a maximal incremental aerobic test. Thirteen male runners performed a maximal incremental aerobic test on an instrumented treadmill to determine their PS. Biomechanical variables were evaluated at the start, mid, and end of each run until volitional exhaustion. The change of running biomechanics with fatigue was similar among the four tested speeds. Duty factor and contact and propulsion times increased with exhaustion (P ≤ 0.004; F ≥ 10.32) while flight time decreased (P = 0.02; F = 6.67) and stride frequency stayed unchanged (P = 0.97; F = 0.00). A decrease in vertical and propulsive peak forces were obtained with exhaustion (P ≤ 0.002; F ≥ 11.52). There was no change in the impact peak with exhaustion (P = 0.41; F = 1.05). For runners showing impact peaks, the number of impact peaks increased (P ≤ 0.04; [Formula: see text] ≥ 6.40) together with the vertical loading rate (P = 0.005; F = 9.61). No changes in total, external, and internal positive mechanical work was reported with exhaustion (P ≥ 0.12; F ≤ 2.32). Results suggest a tendency towards a "smoother" vertical and horizontal running pattern with exhaustion. A smoother running pattern refers to the development of protective adjustments, leading to a reduction of the load applied to the musculoskeletal system at each running step. This transition seemed continuous between the start and end of the running trials and could be adopted by the runners to decrease the muscle force level during the propulsion phase. Despite these changes with exhaustion, there were no changes in either gesture speed (no alteration of stride frequency) or positive mechanical work, advocating that runners unconsciously organize themselves to maintain a constant whole-body mechanical work output.