Aims: Blood biomarkers can improve drug development for Alzheimer’s disease (AD) and its treatment. Neuron-derived extracellular vesicles (NDEVs) in plasma offer a minimally invasive platform for developing novel biomarkers that may be used to monitor the diverse pathogenic processes involved in AD. However, NDEVs comprise only a minor fraction of circulating extracellular vesicles (EVs). Most published studies have leveraged the L1 cell adhesion molecule (L1CAM) for NDEV immunocapture. We aimed to develop and optimize an alternative, highly specific immunoaffinity method to enrich blood NDEVs for biomarker development. Methods: After screening multiple neuronal antigens, we achieved NDEV capture with high affinity and specificity using antibodies against Growth-Associated Protein (GAP) 43 and Neuroligin 3 (NLGN3). The EV identity of the captured material was confirmed by electron microscopy, western blotting, and proteomics. The specificity for neuronal origin was demonstrated by showing enrichment for neuronal markers (proteins, mRNA) and recovery of spiked neuronal EVs. We performed NDEV isolation retrospectively from plasma samples from two cohorts of early AD patients (N = 19 and N = 40) and controls (N = 20 and N = 19) and measured p181-Tau, amyloid-beta (Aβ) 42, brain-derived neurotrophic factor (BDNF), precursor brain-derived neurotrophic factor (proBDNF), glutamate receptor 2 (GluR2), postsynaptic density protein (PSD) 95, GAP43, and syntaxin-1. Results: p181-Tau, Aβ42, and NRGN were elevated in AD samples, whereas proBDNF, GluR2, PSD95, GAP43, and Syntaxin-1 were reduced. Differences for p181-Tau, proBDNF, and GluR2 survived multiple-comparison correction and were correlated with cognitive scores. A model incorporating biomarkers correctly classified 94.7% of AD participants and 61.5% of control participants. The observed differences in NDEVs-associated biomarkers are consistent with previous findings. Conclusion: NDEV isolation by GAP43 and NLGN3 immunocapture offers a robust novel platform for biomarker development in AD, suitable for large-scale validation.