Generative Propaganda
Authors: Madeleine I. G. Daepp, Alejandro Cuevas, Robert Osazuwa Ness et al.
Published: 2025-09-23
arXiv: 2509.19147
Source: arXiv (cs.CY, cs.AI, cs.SI)
Abstract
Generative propaganda is the use of generative artificial intelligence (AI) to shape public opinion. To characterize its use in real-world settings, we conducted interviews with defenders (e.g., factcheckers, journalists, officials) in Taiwan and creators (e.g., influencers, political consultants, advertisers) as well as defenders in India, centering two places characterized by high levels of online propaganda. The term “deepfakes”, we find, exerts outsized discursive power in shaping defenders’ expectations of misuse and, in turn, the interventions that are prioritized. To better characterize the space of generative propaganda, we develop a taxonomy that distinguishes between obvious versus hidden and promotional versus derogatory use. Deception was neither the main driver nor the main impact vector of AI’s use; instead, Indian creators sought to persuade rather than to deceive, often making AI’s use obvious in order to reduce legal and reputational risks, while Taiwan’s defenders saw deception as a subset of broader efforts to distort the prevalence of strategic narratives online. AI was useful and used, however, in producing efficiency gains in communicating across languages and modes, and in evading human and algorithmic detection. Security researchers should reconsider threat models to clearly differentiate deepfakes from promotional and obvious uses, to complement and bolster the social factors that constrain misuse by internal actors, and to counter efficiency gains globally.
Why This Work Matters
This paper delivers the most empirically grounded taxonomy of AI-enabled propaganda to date, drawn from field research rather than adversarial red-teaming. The geographic focus on Taiwan and India — both high-propaganda environments with distinct political contexts — gives the findings cross-regional analytical weight.
Its central correction has direct implications for defensive practice: the deepfake framing dominates defender threat models and drives prioritization of detection tools, yet the study finds that persuasion efficiency (not deception) is the primary AI contribution. This gap between defender assumptions and actual attacker behavior is analytically significant for counter-influence design.
Core Concepts and Contributions
Generative propaganda taxonomy: A 2×2 matrix structured along two axes — obvious vs. hidden (whether AI involvement is disclosed) and promotional vs. derogatory (whether the content supports or attacks a target). This corrects the common conflation of generative propaganda with covert AI-generated disinformation.
Deepfake discourse effect: The term “deepfakes” functions as a cognitive anchor that distorts defenders’ expectations and narrows their countermeasure repertoire. The paper argues this discursive dominance is itself a vulnerability.
Efficiency as the primary threat vector: AI’s main contribution is operational scale and cross-language/cross-modal production speed, not superior deception quality. This shifts the priority from content detection to volume disruption and platform governance.
Social constraint factors: Internal actors (domestic creators) are partly self-limiting due to legal and reputational risks when AI use is obvious — a factor absent from foreign adversarial operations. Counter-influence must therefore differentiate by actor type.
Connections
- Propaganda — core concept note
- Deepfakes — concept directly challenged by this taxonomy
- Computational Propaganda — related IO methodology
- Influence Campaigns — broader operational category
- AI Swarms and Democracy — companion paper on agentic AI IO threat vectors
- Taiwan — primary field research site