Network dysfunction is well established in patients with Alzheimer's disease (AD) and has been shown to be present early in the disease. This is especially interesting in patients with mild cognitive impairment (MCI) since they are more likely to develop AD. In EEG, one type of network analysis is microstates where the EEG is divided into quasi-stable states and these microstates have been linked to networks found with resting state functional MRI. In the current exploratory study, we therefore wanted to explore the changes in microstates in MCI, and AD compared to healthy controls (HC) and whether microstates were able to separate patients with MCI who progressed (pMCI) and those who remained stable (sMCI). EEGs were recorded at baseline for 17 patients with AD, 27 patients with MCI, and 38 older HC and the patients were followed for 3 years. To investigate whole-brain dynamics we extracted different microstate parameters. We found that patients with MCI, and AD had significantly higher occurrence (p-value = 0.028), and coverage (p-value = 0.010) for microstate A compared to HC. However, we did not find any significant systematic deviation of the transition probabilities from randomness for any of the groups. No significant differences were found between pMCI and sMCI but the largest difference in duration was found for microstate D. Microstate A has been linked to the temporal lobes in studies combining EEG and fMRI and the temporal lobes are the most affected by AD pathology in the early stages of the disease. This supports our idea that microstate A may be the first affected microstate in early AD. Even though not significant between pMCI and sMCI, Microstate D has previously been shown to be associated with both frontal and parietal areas as measured with fMRI and may correspond to underlying pathological changes in the progression of MCI to AD. However, larger studies are needed to confirm these findings.