BACKGROUND: Reliable blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) phenotypic biomarkers of Alzheimer's disease (AD) or mild cognitive impairment (MCI) are likely to emerge only from a systematic, quantitative, and aggregate examination of the functional neuroimaging research literature. METHODS: A series of random-effects activation likelihood estimation (ALE) meta-analyses were conducted on studies of episodic memory encoding operations in AD and MCI samples relative to normal controls. ALE analyses were based on a thorough literature search for all task-based functional neuroimaging studies in AD and MCI published up to January 2010. Analyses covered 16 fMRI studies, which yielded 144 distinct foci for ALE meta-analysis. RESULTS: ALE results indicated several regional task-based BOLD consistencies in MCI and AD patients relative to normal control subjects across the aggregate BOLD functional neuroimaging research literature. Patients with AD and those at significant risk (MCI) showed statistically significant consistent activation differences during episodic memory encoding in the medial temporal lobe, specifically parahippocampal gyrus, as well superior frontal gyrus, precuneus, and cuneus, relative to normal control subjects. CONCLUSIONS: ALE consistencies broadly support the presence of frontal compensatory activity, medial temporal lobe activity alteration, and posterior midline "default mode" hyperactivation during episodic memory encoding attempts in the diseased or prospective predisease condition. Taken together, these robust commonalities may form the foundation for a task-based fMRI phenotype of memory encoding in AD.