Background: Neuropathological studies have linked tau aggregates to neuronal loss. To describe the spatial distribution of neurofibrillary tangle pathology in post-mortem tissue, Braak staging has been used. The aim of this study was to examine in vivo associations between tau pathology, quantified with [18F]flortaucipir PET in regions corresponding to Braak stages, and atrophy across the Alzheimer's disease (AD) spectrum. Methods: We included 100 subjects, including 58 amyloid-β positive patients with mild cognitive impairment (MCI, n = 6) or AD dementia (n = 52) and 42 controls with subjective cognitive decline (36% amyloid-β positive). All subjects underwent a dynamic [18F]flortaucipir PET to generate non-displaceable binding potential (BPND) maps. We extracted average [18F]flortaucipir BPND entorhinal, Braak III-IV (limbic) and Braak V-VI (neocortical) regions of interest (ROIs). T1-weighted MRI was used to assess gray matter (GM) volumes. We performed linear regression analyses using [18F]flortaucipir BPND ROIs as independent and GM density (ROI or voxelwise) as dependent variable. Results: In MCI/AD subjects (age [mean ± SD] 65 ± 8 years, MMSE 23 ± 4), [18F]flortaucipir BPND was higher than in controls (age 65 ± 8, MMSE 29 ± 1) across all ROIs (entorhinal 0.06 ± 0.21 vs 0.46 ± 0.25 p < 0.001, Braak III-IV 0.11 ± 0.10 vs 0.46 ± 0.26, p < 0.001, Braak V-VI 0.07 ± 0.07 vs 0.38 ± 0.29, p < 0.001). In MCI/AD, greater [18F]flortaucipir BPND in entorhinal cortex was associated with lower GM density in medial temporal lobe (β - 0.40, p < 0.001). Greater [18F]flortaucipir BPND in ROI Braak III-IV and Braak V-VI was associated with smaller GM density in lateral and inferior temporal, parietal, occipital, and frontal lobes (range standardized βs - 0.30 to - 0.55, p < 0.01), but not in medial temporal lobe (β - 0.22, p 0.07). [18F]Flortaucipir BPND in ROI Braak I-II was not associated with GM density loss anywhere. When quantifying [18F]flortaucipir BPND across brain lobes, we observed both local and distant associations with GM atrophy. In controls, there were no significant associations between [18F]flortaucipir BPND and GM density (standardized βs ranging from - 0.24 to 0.02, all p > 0.05). Conclusions: In MCI/AD patients, [18F]flortaucipir binding in entorhinal, limbic, and neocortical regions was associated with cortical atrophy.