Not “just a language model”: AI’s ideological functions

Abstract This essay examines how artificial intelligence systems perform ideological functions that extend far beyond technical bias or governance protocols. Moving past ethical frameworks that treat AI as neutral tools requiring debiasing, it maps five coordinates through which AI amplifies and transforms ideology: infrastructure, where oligopolistic control is naturalized as democratic progress; normalization, where statistical reproduction of social hierarchies appears as procedural objectivity; subjectification, where users are interpellated through predictive intimacy and algorithmic mirroring; opacity, where contradiction is managed through affective saturation and epistemic overwhelm; and pastiche, where coherent fragments circulate without anchoring, enabling cynical participation in systems users know to be artificial. These operations reveal AI not as mere language modeling but as a remodeling of structured intelligibility, depoliticizing domination by formatting it as technical necessity. The essay traces how contemporary AI inherits and mutates classical ideological functions—from Marx’s camera obscura to Althusser’s interpellation to Debord’s spectacle—although generating new mechanisms of control through rhythm, mimicry, and synthetic coherence. It concludes by identifying three counter-pressure vectors: infrastructural reclamation, generative misuse, and redeployment outward—tactical approaches for contesting AI’s ideological capture without mythologizing technological resistance.