To investigate whether progression of cognitive impairment has an effect on connectivity gradients, and to examine the differences in gradient changes across brain regions due to the progression of cognitive impairment. Methods: The resting-state fMRI data from 40 normal controls (NC group), 42 patients with early mild cognitive impairment (MCI) (EMCI group), 39 patients with late MCI (LMCI group), and 40 patients with Alzheimer's disease (AD) (AD group) were collected. A nonlinear decomposition method was used to calculate the linker gradient from the functional connectivity data of each subject group and to compare the differences in gradients between subject groups at different stages. Results: Progression of cognitive impairment leads to an increase in the number of brain networks that change in the connectivity pattern of the connectivity gradient and a significant decrease in gradient scores, with the most significant change in the first gradient occurring during the EMCI stage and the most significant change in the second gradient occurring during the AD stage, and with the first gradient explaining the largest difference in gradient change. Conclusion: The progression of cognitive impairment impacts the connectivity gradient. Compared to the discrete network approach, the connectivity gradient preserves the connections among the subnetworks and responds to the macroscopic connectivity patterns of the whole brain.