Abstract: Based on nine cases from national, provincial, county-level, and commercial platforms, this article examines how algorithms inherit and transform traditional editorial gatekeeping through boundary demarcation, identity construction, and network formation. Findings reveal that classification, indexing, and association constitute technological gatekeeping, giving rise to three types of operative memory: archive-dominated, narrative-dominated, and mixed—which oscillates between computability and narrativity, reconstructing the spatiotemporal order of news memory. This nested logic intensifies risks of memory commodification, cognitive closure, and historical simplification in the short term, and may lead to narrative fragmentation, memory stratification, and algorithmic filtering in the long term. The article argues that good algorithmic editing should seek dynamic equilibrium across evidentiary preservation, narrative pluralism, cognitive fairness, and temporal integrity, and proposes “reflective editing” as a critical framework for scrutinizing implicit power operations within technological gatekeeping. 

Key words: News digital memory　Technological gatekeeping　Operative memory　Algorithmic editing　Reflective editing