Despite strong evidence from animal models of Alzheimer's disease (AD) supporting aerobic exercise as a disease-modifying treatment for AD, human mechanistic studies are limited with mixed findings. The objective of this pilot randomized controlled trial was to examine the effects of 6-month aerobic exercise on hippocampal volume, temporal meta-regions of interest (ROI) cortical thickness, white matter hyperintensity (WMH) volume, and network failure quotient (NFQ), measured with MRI, in community-dwelling older adults with AD dementia. Additionally, the relationships between 6- and 12-month changes in MRI biomarkers and the AD Assessment Scale-Cognition (ADAS-Cog) were examined. Sixty participants were randomized, but one was excluded because baseline MRI failed quality control: 38 randomized to cycling and 21 to stretching. The intervention was moderate-intensity cycling for 20–50 mins, three times a week for 6 months. Control was low-intensity stretching. The study outcomes include hippocampal volume, temporal meta-ROI cortical thickness, WMH volume, and NFQ. Outcomes were measured at baseline, 6 months, and 12 months. The sample averaged 77.3 ± 6.3 years old with 15.6 ± 2.9 years of education and 53% men. Both groups experienced significant declines over 6 months in hippocampal volume (2.64% in cycling vs. 2.89% in stretching) and temporal meta-ROI cortical thickness (0.94 vs. 1.54%), and over 12 months in hippocampal volume (4.47 vs. 3.84%) and temporal meta-ROI cortical thickness (2.27 vs. 1.79%). These declines did not differ between groups. WMH volume increased significantly with the cycling group increasing less (10.9%) than stretching (24.5%) over 6 months (