AI-Generated Propaganda

A · DEFINITION

What this category means

Organized persuasive communication in which generative AI creates, substantially transforms, localizes, or mass-produces text, images, audio, video, memes, or synthetic documents.

Outside its scope

Not every synthetic artifact is propaganda. The category requires an organized political, ideological, military, commercial, or institutional objective.

Evidence basis^[1]

B · WHY IT MATTERS

Strategic and public-interest significance

Propaganda depends on repetition, timing, identity cues, and distribution as much as technical realism. AI expands the volume and variety of available assets and can also create a liar’s dividend in which authentic evidence is dismissed as synthetic.

Primary AI role: Tool
Unit of influence: group / population
Degree of autonomy: Low to moderate
Evidence maturity: Documented current use

Evidence basis^{[1, 2, 3]}

C · HOW AI CHANGES IT

What changes compared with pre-AI practice

LLMs can rewrite a core narrative in many languages and styles, while image, audio, and video models create emotionally salient evidence-like assets. Human-machine teams can match human propaganda in controlled persuasion studies, but real-world behavioral effects remain hard to establish.

Evidence basis^{[1, 3]}

D · CAPABILITY STATUS

Separate current evidence from prospective risk

Confirmed real-world use

State-aligned and commercial operations have used generative AI for articles, comments, translation, synthetic presenters, and cloned voices.

Demonstrated technical capability

Controlled experiments find AI-generated propaganda can be about as persuasive as human-written material, especially with human curation.

Plausible near-term development

Faster multimodal production and cross-platform adaptation are likely to expand.

Speculative or unsupported claims

No evidence supports universal mind control or reliable conversion of exposure into action.

Evidence basis^{[1, 2, 3, 4, 5]}

E · KEY MECHANISMS

Conceptual mechanisms—not procedures

01

High-volume linguistic variation and localization.

02

Synthetic media that exploits emotional intensity and information vacuums.

03

Human curation that selects plausible outputs and removes obvious errors.

04

Distribution through inauthentic accounts, deceptive sites, messaging channels, or robocalls.

Evidence basis^{[1, 3]}

F · EVIDENCE AND EXAMPLES

What occurred, what is known, and what remains unknown

Reach, engagement, and visibility are not treated as proof of persuasion or behavior change.

Compare every qualified case across the taxonomy

CopyCop / DC Weekly^{[1, 3]}

What occurred: A Russia-linked network used generative rewriting to expand output across deceptive news properties.
Evidence status: AI use was confirmed through text artifacts and longitudinal analysis.
Measured or documented effect: Publication volume and narrative breadth increased without an observed loss of survey-rated persuasiveness.
What remains unknown: Long-term changes in voting or foreign-policy attitudes cannot be isolated from this evidence.

New Hampshire primary robocall^{[1, 4]}

What occurred: A cloned voice resembling President Biden told voters not to participate in the January 2024 primary.
Evidence status: The use of voice cloning and the responsible political consultant were confirmed by enforcement actions.
Measured or documented effect: Roughly 9,600 calls prompted substantial regulatory and legal action.
What remains unknown: The number of voters actually deterred remains unquantified.

Wolf News synthetic presenters^{[1, 5]}

What occurred: A pro-China network used commercial synthetic avatars to deliver English-language political messages.
Evidence status: AI-generated presenters and the distribution network were documented.
Measured or documented effect: The videos generally received very low engagement.
What remains unknown: It is unclear whether this was a capability test or a fully resourced operation.

G · RISKS AND FAILURE MODES

Malicious-use risks and reasons the capability may fail

Primary risks

Synthetic evidence can provoke action before verification.
Awareness of deepfakes can be exploited to deny authentic evidence.
Cheap output can overwhelm fact-checking and moderation capacity.

Evidence basis^{[1, 2, 4, 5]}

Limits and failure modes

Distribution and trust remain harder than generation.
Technical realism is context dependent and many campaigns receive little organic attention.

Evidence basis^{[1, 5]}

H · DETECTION AND DEFENSIVE INDICATORS

Signals are suggestive, not automatic proof

False-positive warning: No single detector score, writing style, profile image artifact, posting pattern, or political similarity should be used alone to accuse a person or organization.

Unverified media released into a breaking-news vacuum warrants source-first verification, not automatic rejection.
Shared infrastructure and coordinated timing are often more informative than detector scores.

Evidence basis^{[1, 3, 4]}

I · GOVERNANCE AND SAFEGUARDS

Layered controls, oversight, and accountability

Use authenticated institutional channels and rapid crisis verification.
Adopt provenance while planning for metadata loss and spoofing.
Prioritize narrative- and source-level triage over endless item-by-item debunking.

Evidence basis^{[1, 4]}

Related cross-category safeguards

The resilience guide compares these controls with their limits and evidence context across the full taxonomy.

Media literacy and pre-bunking Source verification and crisis procedures Provenance and authenticity signals Institutional communication and trust repair

Legal conclusions depend on jurisdiction and facts; this page summarizes the corresponding report and is not legal advice.

J · RESEARCH GAPS

Questions the evidence does not yet resolve

Longitudinal behavioral effects outside survey settings.
How labels affect true and false content across different audiences.
Real-world comparative effectiveness of synthetic media modalities.

Evidence basis^{[1, 2, 3]}

Compare this category’s questions across the research agenda

K · SOURCES

Traceable source list

The commissioned report is the organizing source. External records below are the principal sources retained for the public synthesis; source quality varies by type and is labelled.

English and Spanish editions are published from the same structured record. Bilingual parity is validated for every release; source titles may remain in their original publication language.

Browse the cross-category case register Browse the cross-category evidence register Compare the terminology used across categories Compare cross-category resilience safeguards Review the cross-category open research agenda

Commissioned research report The Architecture of Artificial Influence: A Comprehensive Report on AI-Generated Propaganda
Evidence links: 13
Peer-reviewed research How persuasive is AI-generated propaganda?
Evidence links: 4
Peer-reviewed research Generative propaganda: Evidence of AI’s impact from a state-backed disinformation campaign
Evidence links: 7
Official enforcement record FCC 24-59 enforcement document
Evidence links: 5
Secondary threat reporting Deepfake news anchors spread Chinese propaganda on social media
Evidence links: 4