Biomarkers in Alzheimer's disease (AD) have the potential to allow early and more accurate diagnosis, predict disease progression, stratify individuals and track response to candidate therapies in drug trials. The first fluid biomarkers reflecting aspects of AD neuropathology were identified in cerebrospinal fluid (CSF) in the 1990s. Three CSF biomarkers (amyloid-β 1-42, total tau and phospho-tau) have consistently been shown to have diagnostic utility and are incorporated into the new diagnostic criteria for AD. These markers have also been shown in longitudinal studies to predict conversion of mild cognitive impairment to AD. However, a key issue with the use of CSF biomarkers as a screening test is the invasiveness of lumbar puncture. Over the last 20 years there has been an active quest for blood biomarkers, which could be easily acquired and tested repeatedly throughout the disease course. One approach to identifying such markers is to attempt to measure candidates that have already been identified in CSF. Until recently, this approach has been limited by assay sensitivity, but newer platforms now allow single molecule-level detection. Another approach is identification of candidates in large multiplex panels that allow for multiple analytes to be quantified in parallel. While both approaches show promise, to date no blood-based biomarker or combination of biomarkers has sufficient predictive value to have utility in clinical practice. In this review, an overview of promising blood protein candidates is provided, and the challenges of validating and converting these into practicable tests are discussed.