In addition to the deposition of β-amyloid plaques, neurofibrillary tangles composed of aggregated hyperphosphorylated tau are one of the pathological hallmarks of Alzheimer's disease and other neurodegenerative disorders. Until now, our understanding about the natural history and topography of tau deposition has only been based on post-mortem and cerebrospinal fluid studies, and evidence continues to implicate tau as a central driver of downstream neurodegenerative processes and cognitive decline. Recently, it has become possible to assess the regional distribution and severity of tau burden in vivo with the development of novel radiotracers for positron emission tomography (PET) imaging. In this article, we provide a comprehensive discussion of tau pathophysiology, its quantification with novel PET radiotracers, as well as a systematic review of tau PET imaging in normal aging and various dementia conditions: mild cognitive impairment, Alzheimer's disease, frontotemporal dementia, progressive supranuclear palsy, and Lewy body dementia. We discuss the main findings in relation to group differences, clinical-cognitive correlations of tau PET, and multi-modal relationships among tau PET and other pathological markers. Collectively, the small but growing literature of tau PET has yielded consistent anatomical patterns of tau accumulation that recapitulate post-mortem distribution of neurofibrillary tangles which correlate with cognitive functions and other markers of pathology. In general, AD is characterised by increased tracer retention in the inferior temporal lobe, extending into the frontal and parietal regions in more severe cases. It is also noted that the spatial topography of tau accumulation is markedly distinct to that of amyloid burden in aging and AD. Tau PET imaging has also revealed characteristic spatial patterns among various non-AD tauopathies, supporting its potential role for differential diagnosis. Finally, we propose novel directions for future tau research, including (a) longitudinal imaging in preclinical dementia, (b) multi-modal mapping of tau pathology onto other pathological processes such as neuroinflammation, and (c) the need for more validation studies against post-mortem samples of the same subjects.