A multicenter prospective cohort trial evaluated a fiber-optic Raman spectroscopy AI system, SPECTRA IMDx, against routine white-light endoscopy for identifying high-risk gastric neoplasia during active endoscopic examination. In the full set of 238 gastric lesions from 198 patients, the Raman AI system improved diagnostic accuracy, sensitivity, and specificity. Reported performance included 90.0% sensitivity versus 58.0% for routine white-light endoscopy and specificity of 81.4% versus 58.5%. Paired analysis against combined white-light endoscopy plus blue laser imaging showed similar diagnostic accuracy between modalities in that subset. The study positions Raman spectroscopy with AI as a potential in-procedure diagnostic upgrade for endoscopists, with histopathology serving as the reference standard. Implementation will hinge on throughput, integration with existing endoscopy workflows, and validation across broader sites.