The β-amyloid peptide (Aβ) is the main constituent of senile plaques, a typical hallmark of Alzheimer’s disease (AD). Monomeric Aβ is generated through sequential processing of the amyloid precursor protein (APP), with a final step involving γ-secretase activity. In AD, Aβ monomers assemble in oligomers and ultimately fibrils depositing in plaques. Importantly, Aβ toxicity appears related to its soluble oligomeric intermediates. In particular, recombinant Aβ studies described Aβ hexamers as critical oligomeric nuclei. We recently identified hexameric Aβ assemblies in a cellular model, and revealed their ability to enhance recombinant Aβ aggregation in vitro. Here, we assessed the contribution of similar hexameric-like Aβ assemblies to the development of amyloid pathology. We report their early presence in both transgenic mice brains exhibiting human Aβ pathology and cerebrospinal fluid of AD patients, suggesting hexameric Aβ as a putative novel AD biomarker. Using isolated cell-derived hexameric Aβ, we report the potential of these assemblies to seed other human Aβ species, resulting in neuronal toxicity in vitro and amyloid deposition aggravation in vivo. In order to identify key contributors to their formation in a cellular context, we investigated the role of presenilin-1 (PS1) and presenilin-2 (PS2) in the formation of hexameric-like Aβ assemblies. As catalytic subunits of the γ-secretase complex, PS1 and PS2 can differentially participate in Aβ generation. Using CRISPR-Cas9-modified neuronal-like cell lines knockdown for each of the two presenilins, we present experimental evidence suggesting a direct link between the PS2-dependent pathway and the release of hexameric-like Aβ assemblies in extracellular vesicles.