PURPOSE: Vessel-encoded arterial spin labeling (VEASL) enables simultaneous, non-contrast imaging of multiple vascular territories that is useful for differential diagnosis and treatment monitoring of cerebrovascular diseases. However, the existing encoding methods are signal-to-noise ratio (SNR) inefficient. METHODS: We developed a MOdulation-Guided ENcoding (MOGEN) scheme that directly exploits the inversion spatial modulation profile to obtain SNR-efficient encoding matrix. Simulations, phantom tests, and healthy volunteer scans were performed to demonstrate its feasibility in multiple application scenarios. RESULTS: Simulation studies demonstrated that MOGEN achieves significantly higher theoretical SNR efficiency than previous methods for both four- and six-artery configurations. In healthy volunteers, MOGEN improved in vivo SNR by approximately 15% and provided more robust vessel decoding, particularly when the spatial modulation deviated from a cosine profile. In patients with Moyamoya disease, MOGEN enabled reliable visualization of collateral pathways even when scan time was reduced to ˜5 min for six arteries. Furthermore, by considering vessel size with multi-voxel vessel representation, MOGEN enhanced single-artery selectivity in vessel-encoded angiography. We also demonstrated that a straightforward approach of off-resonance correction for VEASL at ultra-high field was feasible by using MOGEN. CONCLUSION: MOGEN offered several benefits for VEASL, including high SNR efficiency, flexible spatial modulation and PCASL parameters selection, vessel size consideration, and straightforward off-resonance correction, thereby substantially improving robustness and usability of VEASL across various applications.