Novel diagnostic criteria for Alzheimer's disease (AD) incorporate biomarkers, but their maturity for implementation in clinical practice at the prodromal stage (mild cognitive impairment [MCI]) is unclear. Here, we evaluate cerebrospinal fluid (CSF) β-amyloid42 (Aβ42), total tau, and phosphorylated tau in the light of a 5-phase framework for biomarker development. Ample evidence is available for phase 1 (identifying useful leads) and phase 2 (assessing the accuracy for AD dementia versus controls) for CSF biomarkers. Phase 3 (utility in MCI) is partially achieved. In cohorts with long follow-up time, CSF Aβ42, total tau, and phosphorylated tau have high diagnostic accuracy for MCI due to AD. Phase 4 (performance in real world) is ongoing, and phase 5 studies (quantify impact and costs) are to come. Our results highlight priorities to pursue and to enable the proper use of CSF biomarkers in the clinic. Priorities are to reduce measurement variability by introduction of fully automated assay systems; to increase diagnostic specificity toward non-AD neurocognitive diseases at the MCI stage; and to clarify the role of CSF biomarkers versus other biomarker modalities in clinical practice and in design of clinical trials. These efforts are currently ongoing.