The preclinical phase of Alzheimer's disease (AD) occurs years, possibly decades, before the onset of clinical symptoms. Being able to detect the very earliest stages of AD is critical to improving understanding of AD biology, and identifying individuals at greatest risk of developing clinical symptoms with a view to treating AD pathophysiology before irreversible neurodegeneration occurs. Studies of dominantly inherited AD families and longitudinal studies of sporadic AD have contributed to knowledge of the earliest AD biomarkers. Here we appraise this evidence before reviewing novel, particularly fluid, biomarkers that may provide insights into AD pathogenesis and relate these to existing hypothetical disease models.