This study aims to address the limitations of traditional news reporting in terms of spatial presentation dimensions and user cognitive participation, especially the lack of context restoration and information perception discretization caused by two-dimensional media, and proposes an immersive news framework based on augmented reality (AR) spatial computing. By building an AR news system that integrates SLAM positioning, photon mapping rendering, and multi-channel interaction (the hardware uses Hololens 2, and the software architecture includes a 12-layer neural network-driven semantic understanding module), a control experiment is designed to collect behavioral data from 90 subjects: The experimental group explores the dynamic evolution of the Syrian war events by operating the three-dimensional timeline through gestures, while the control group reads a two-dimensional long text on the same topic and uses multimodal data capture technology to record the user's spatial movement trajectory, interaction hot spot distribution and pupil diameter changes. Experimental data shows that the AR group has an accuracy rate of 92% in reconstructing event causal relationships in the information integration index. Behavioral path analysis shows that users trigger an average of 3.36 perspective zooms and 7.9 space–time layer overlay operations. Physiological measurements show that the AR group has a pupil diameter peak of 4.8 mm at key narrative nodes, providing a quantitative threshold for the narrative density design of immersive news.