Answers are grounded in this paper and cite their sources.
Generative large language models (LLMs) increasingly shape public discourse, yet their role in framing reality remains underexplored. This study introduces the concept of AI super-frames, high-level narrative patterns embedded in synthetic news output, and examines how ChatGPT and DeepSeek construct and circulate such frames. Drawing on 144 human- and AI-generated news articles derived from CNN and CCTV headlines, the study combines NVivo-assisted thematic coding, mixed-method framing analysis, and Critical Discourse Analysis. Findings identify six dominant AI super-frames: Conflict & Competition, Security & Threat, Moral Responsibility, Technological Solutionism, Economic Progress, and Human Progress, alongside tendencies toward artificial neutrality and procedural legitimacy. Despite identical prompts, ChatGPT amplified U.S. security concerns and geopolitical rivalry, whereas DeepSeek foregrounded technological progress and China’s global ambitions. These findings extend framing theory by demonstrating that framing agency can emerge from algorithmic probability rather than deliberate human intent, highlighting implications for digital literacy, journalism, and AI governance.
No takes yet. Share an insight, caveat, or question.
Nowacki et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: