Educational companion dossier · Fact, interpretation, lived experience, clinical education, fiction, and mechanics are labeled separately. Scope & safety
PsychologicalWar.orgHistory before hype.

AI PSYOPS TAXONOMY · CATEGORY 02

AI-Generated Propaganda

Generative AI creates, transforms, localizes, or mass-produces propaganda across text, image, audio, video, memes, websites, or synthetic documents.

Primary level: Tool Documented current use Claim AIP-02-A01
Read evidence assessment All 12 categories

CURRENT EVIDENCE ASSESSMENT

AI-generated propaganda is documented and experimentally persuasive in bounded settings; field effects vary and remain difficult to attribute.

Stable claim identifierAIP-02-A01
Claim stageassessment
Currentness reviewed2026-07-27T20:15:00Z

DEPLOYMENT

Documented

Text, image, audio, and synthetic presenter use is documented in multiple public cases.

AUTONOMY

Usually human-directed

Humans still select narratives, timing, channels, and strategic objectives in documented operations.

PERSISTENCE

Sustained production is feasible

AI lowers content-production friction, but sustained audience penetration still depends on infrastructure and distribution.

PROFILING ACCURACY

Optional rather than inherent

Propaganda can be mass-produced without profiling; tailored variants inherit the limits of personalization evidence.

MEASURED EFFECT

Short-term attitude effects demonstrated; field behavior unresolved

Controlled studies find persuasive potential, but operational outcomes and durable behavior are not established by content volume.

Assessment basis

Assessment combines the exact owner-supplied category report with the bounded primary, official, platform, and peer-reviewed sources listed for this category. Dimensions are evaluated separately to prevent documented output from being mistaken for autonomy or effect. WIP.51 adds official and academic closure for the separate Slovakia audio incident while retaining an unresolved effect boundary.

What would change this assessment

Reclassify upward only if independent field studies connect authenticated AI content exposure to durable attitude or behavior change.

Prohibited inference

Do not infer strategic effect, universal deployment, or individual psychological state from this assessment.

A · DEFINITION

What this category means sources

Definition

AI-generated propaganda is organized influence content for which generative systems create or substantially transform the media itself. It includes synthetic text, images, audio, video, avatars, memes, documents, and cloned news-style content.

Outside this category

Not every AI-assisted communication is propaganda. Transparent public information, disclosed satire, ordinary creative work, and benign translation fall outside this category unless they are organized to advance an influence objective through misleading or manipulative presentation.

B · SIGNIFICANCE

Why it matters sources

Generative systems reduce the cost of producing many plausible variants in multiple languages and formats. The risk is not limited to believing a single fake: a high-volume synthetic environment can exhaust verification systems, create false consensus, and weaken confidence in genuine evidence.

C · CHANGE FROM PRE-AI PRACTICE

How AI changes the phenomenon sources

AI turns content production from a labor bottleneck into a scalable service. It can rapidly rewrite, localize, summarize, or combine material. Yet realism is not the same as persuasiveness, and the report repeatedly separates production, distribution, exposure, and behavioral effect.

D · CAPABILITY STATUS

Separate evidence from projection sources

DOCUMENTED

Confirmed real-world use

Publicly documented influence operations have used generative text, synthetic presenters, cloned voices, and altered media in political and geopolitical contexts.

DEMONSTRATED

Demonstrated technical capability

Controlled research finds that AI-generated persuasive text can perform similarly to human-written material under limited conditions.

EMERGING

Plausible near-term development

Cross-modal pipelines may combine text, image, audio, and video with rapid localization and automated variation.

UNCERTAIN

Unsupported or unproven

No reliable evidence establishes that synthetic media consistently produces durable behavior change or “controls” a population.

E · KEY MECHANISMS

Conceptual mechanisms — not an operating procedure sources

Safety transformation: these descriptions identify system functions at a high level. Procedural steps, target criteria, scripts, evasion methods, and deployment workflows are intentionally excluded.

  1. Mass production of rhetorically varied text and pseudo-news content.
  2. Synthetic images designed for emotional salience or false evidentiary value.
  3. Voice cloning, synthetic presenters, and altered audiovisual statements.
  4. Rapid translation and local adaptation across languages and platforms.
  5. Hybrid human-machine editing that combines strategic oversight with automated production.

F · EVIDENCE & EXAMPLES

What is known, measured, and still unknown sources

REACH IS NOT EFFECT. Publication, impressions, engagement, virality, or media attention do not by themselves establish persuasion or behavioral change.

Example 1 · Claim AIP-02-E01

Doppelganger and DC Weekly

Confirmed AI-assisted production in documented networks

What occurred
Cloned media brands and pseudo-news properties distributed rewritten and localized political narratives.
What is confirmed
Generative rewriting and coordinated deceptive infrastructure are documented.
Effect measured
Publication volume and narrative breadth increased.
What remains unknown
The causal effect on beliefs, voting, or offline behavior remains uncertain.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.

Permanent claim linkCorrection process

Inspect the 20-stage evidence boundary
Artifact Or Event Existence
SUPPORTED_BY_LINKED_SOURCE
Content Status
BOUNDED_DESCRIPTION_SUPPORTED
Coordination
SOURCE_DEPENDENT_OR_UNRESOLVED
Actor Identity
SOURCE_DEPENDENT_OR_UNRESOLVED
Sponsorship Or Direction
SOURCE_DEPENDENT_OR_UNRESOLVED
Intent
SOURCE_DEPENDENT_OR_UNRESOLVED
Output
DOCUMENTED_OR_DESCRIBED_IN_LINKED_SOURCE
Distribution
PARTIAL_OR_SOURCE_DEPENDENT
Availability
PARTIAL_OR_SOURCE_DEPENDENT
Reach
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Exposure
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Attention
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Recall
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Comprehension
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Credibility
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Belief Or Attitude
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Intention
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Behavior
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Operational Outcome
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Strategic Effect
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED

Competing explanations: The observed artifact or action may have depended on human direction, pre-existing networks, platform incentives, ordinary automation, non-AI methods, or unrelated contextual factors.

Affected-person/community evidence: Direct affected-person or affected-community evidence was not independently retrieved for this bounded claim unless explicitly stated in the linked source scope.

Rights and privacy: Expression, provenance, election integrity, crisis communication, and anti-censorship safeguards must be balanced.

Reopening trigger: Reopen this claim when a primary, official, adjudicative, peer-reviewed, affected-person, or affected-community source materially changes identity, attribution, autonomy, distribution, effect, rights, or currentness.

Linked sources: SRC-04-GERMAN-FO-DOPPELGANGER German Federal Foreign Office SRC-06-RECORDED-FUTURE-COPYCOP Recorded Future / Insikt Group
Example 2 · Claim AIP-02-E02

Spamouflage / “Wolf News”

Confirmed synthetic presenters and coordinated distribution

What occurred
A state-aligned network used synthetic news-style presenters and multilingual content.
What is confirmed
The use of synthetic avatars and coordinated inauthentic distribution is documented.
Effect measured
The content achieved visibility but often limited authentic engagement.
What remains unknown
Low engagement cannot prove zero exposure or zero effect, and high visibility would not prove persuasion.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.

Permanent claim linkCorrection process

Inspect the 20-stage evidence boundary
Artifact Or Event Existence
SUPPORTED_BY_LINKED_SOURCE
Content Status
BOUNDED_DESCRIPTION_SUPPORTED
Coordination
SOURCE_DEPENDENT_OR_UNRESOLVED
Actor Identity
SOURCE_DEPENDENT_OR_UNRESOLVED
Sponsorship Or Direction
SOURCE_DEPENDENT_OR_UNRESOLVED
Intent
SOURCE_DEPENDENT_OR_UNRESOLVED
Output
DOCUMENTED_OR_DESCRIBED_IN_LINKED_SOURCE
Distribution
PARTIAL_OR_SOURCE_DEPENDENT
Availability
PARTIAL_OR_SOURCE_DEPENDENT
Reach
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Exposure
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Attention
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Recall
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Comprehension
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Credibility
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Belief Or Attitude
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Intention
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Behavior
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Operational Outcome
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Strategic Effect
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED

Competing explanations: The observed artifact or action may have depended on human direction, pre-existing networks, platform incentives, ordinary automation, non-AI methods, or unrelated contextual factors.

Affected-person/community evidence: Direct affected-person or affected-community evidence was not independently retrieved for this bounded claim unless explicitly stated in the linked source scope.

Rights and privacy: Expression, provenance, election integrity, crisis communication, and anti-censorship safeguards must be balanced.

Reopening trigger: Reopen this claim when a primary, official, adjudicative, peer-reviewed, affected-person, or affected-community source materially changes identity, attribution, autonomy, distribution, effect, rights, or currentness.

Linked sources: SRC-07-GRAPHIKA-WOLF-NEWS Graphika SRC-08-META-SPAMOUFLAGE
Example 3 · Claim AIP-02-E03

Election-period synthetic audio: Slovakia reporting and the New Hampshire robocall

New Hampshire AI use confirmed by enforcement record; Slovakia attribution retained as an unresolved report lead

What occurred
The source report discusses election-period synthetic audio incidents. The FCC record independently confirms the New Hampshire voice-cloning robocall; this review did not independently close the separate Slovakia audio claim.
What is confirmed
AI voice cloning and election misinformation are confirmed for the New Hampshire robocalls. The Slovakia example remains bounded to the owner-supplied report pending direct authoritative closure.
Effect measured
Distribution and an enforcement response are documented for New Hampshire; persuasion or turnout change is not established.
What remains unknown
Recipient belief, behavior, and any effect on turnout remain unknown; the separate Slovakia incident still requires direct source verification.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.

Permanent claim linkCorrection process

Inspect the 20-stage evidence boundary
Artifact Or Event Existence
SUPPORTED_FOR_NEW_HAMPSHIRE_AND_SLOVAKIA
Content Status
SYNTHETIC_OR_FAKE_AUDIO_SUPPORTED_IN_LINKED_SOURCES
Coordination
PARTIAL_NEW_HAMPSHIRE_ENFORCEMENT_SLOVAKIA_UNRESOLVED
Actor Identity
PARTIAL_NEW_HAMPSHIRE_SLOVAKIA_NOT_RETRIEVED
Sponsorship Or Direction
PARTIAL_OR_NOT_RETRIEVED
Intent
ELECTION_INFLUENCE_CONTEXT_SUPPORTED_EXACT_INTENT_NOT_UNIFORMLY_ADJUDICATED
Output
SYNTHETIC_AUDIO_DOCUMENTED
Distribution
DOCUMENTED_WITHOUT_COMPLETE_PATH
Availability
DOCUMENTED
Reach
NOT_RETRIEVED_UNIQUE_COUNT
Exposure
NOT_RETRIEVED
Attention
PARTIAL_PUBLIC_ATTENTION_NOT_AUDIENCE_MEASURE
Recall
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Comprehension
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Credibility
NOT_ESTABLISHED_UNLESS_EXPLICITLY_MEASURED
Belief Or Attitude
NOT_ESTABLISHED
Intention
NOT_ESTABLISHED
Behavior
NOT_ESTABLISHED
Operational Outcome
ENFORCEMENT_RESPONSE_DOCUMENTED_FOR_NEW_HAMPSHIRE
Strategic Effect
NOT_ESTABLISHED

Competing explanations: The observed artifact or action may have depended on human direction, pre-existing networks, platform incentives, ordinary automation, non-AI methods, or unrelated contextual factors.

Affected-person/community evidence: Direct affected-person or affected-community evidence was not independently retrieved for this bounded claim unless explicitly stated in the linked source scope.

Rights and privacy: Expression, provenance, election integrity, crisis communication, and anti-censorship safeguards must be balanced.

Reopening trigger: Reopen this claim when a primary, official, adjudicative, peer-reviewed, affected-person, or affected-community source materially changes identity, attribution, autonomy, distribution, effect, rights, or currentness.

Linked sources: SRC-17-FCC-NH-DEEPFAKE-ROBOCALL Federal Communications Commission

G · RISKS & FAILURE MODES

Potential harms and reasons the capability may fail sources

Risks

  • Synthetic documents or recordings can create false evidence during crises.
  • Content flooding can overwhelm fact-checking and institutional response capacity.
  • The liar’s dividend allows authentic material to be dismissed as fabricated.
  • Detectors can falsely accuse human authors or miss newer generation methods.
  • Rapid replication can preserve a narrative after individual accounts or domains are removed.

Limitations and failure modes

  • AI output may contain factual, cultural, visual, or temporal inconsistencies.
  • Audience effects depend on source credibility, identity, timing, and prior belief—not realism alone.
  • Detection and generation systems evolve in an adversarial cycle.
  • Laboratory attitude shifts may not reproduce in distracted, real-world media environments.

H · DETECTION & DEFENSIVE INDICATORS

Signals for investigation, not automatic verdicts sources

Indicator rule: unless the source report supports a stronger conclusion, each signal below is suggestive rather than conclusive. Multiple independent signals and contextual evidence are required.

  • Near-simultaneous publication of many semantically similar but linguistically varied stories may justify coordination review.
  • Cloned branding, domain imitation, and fake bylines are stronger evidence than an “AI-sounding” sentence.
  • Missing or broken provenance may increase uncertainty but does not prove fabrication.
  • Audio or video artifacts can support forensic review but should not be used as a single decisive test.

I · GOVERNANCE & SAFEGUARDS

Accountability, transparency, and human protection sources

Cryptographically sign authoritative communications and maintain independent verification channels.

Use layered newsroom and crisis verification procedures before repeating sensational media.

Design platform responses around coordinated behavior, infrastructure, and velocity—not only content classification.

Label synthetic media carefully while avoiding claims that unlabeled media is therefore authentic.

Prepare public institutions to confirm genuine statements quickly without over-amplifying false material.

J · RESEARCH GAPS

Questions the evidence does not yet close sources

  • Longitudinal effects of repeated synthetic media exposure.
  • Whether disclosure labels reduce harm or create false confidence in unlabeled content.
  • How multimodal propaganda changes trust across different cultural and media systems.
  • Field evidence connecting exposure to voting, mobilization, or violence.

K · SPECIALIST REVIEW PACKET

Prepared for independent review; no disposition recorded

PacketAIP-02-SPECIALIST-REVIEW-PACKET
DispositionPENDING
Completed dispositions0
Prepared2026-07-27T20:15:00Z

Requested reviewer domains

  • propaganda studies
  • platform integrity
  • political communication
  • media forensics

Questions for reviewers

  1. Are the Slovakia and New Hampshire incidents kept distinct in artifact, attribution, enforcement, and effect?
  2. Are laboratory persuasion results described without implying field effectiveness?
  3. Does the source set adequately represent affected audiences and crisis contexts?

Unresolved questions

  • Longitudinal effects of repeated synthetic media exposure.
  • Whether disclosure labels reduce harm or create false confidence in unlabeled content.
  • How multimodal propaganda changes trust across different cultural and media systems.
  • Field evidence connecting exposure to voting, mobilization, or violence.

Correction and reopening

Correction trigger: Reclassify upward only if independent field studies connect authenticated AI content exposure to durable attitude or behavior change.

Reopening trigger: Reopen this claim when a primary, official, adjudicative, peer-reviewed, affected-person, or affected-community source materially changes identity, attribution, autonomy, distribution, effect, rights, or currentness.

Prepared packet is not completed specialist review, factual certification, legal advice, clinical review, accessibility certification, publication approval, or production authority.

REAL-WORLD INTERPRETIVE

M · SOURCES & REVIEW STATUS

Exact owner report, claim register, and reviewed sources

  1. AI-Generated Propaganda Owner-supplied report: AI Propaganda Research Report.md · 57,697 bytes · SHA-256 c58150bf3ec9bcbd3312190a8119bf00c7a973b3f809b56301edb9c8e2e7a212

    Owner-supplied interdisciplinary research synthesis; exact source preserved in protected durable memory. External specialist review remains pending.

Claim-specific reviewed sources

  1. OpenAI · 2024-05-30 · Authoritative first-party platform disclosure

    Supports
    Documents five disrupted covert influence operations using OpenAI services and reports no meaningful increase in audience engagement or reach attributable to those services as of the publication date.
    Does not establish
    Does not measure all exposure, belief change, behavior, or strategic effect; platform visibility is necessarily partial.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  2. German Federal Foreign Office · 2024-01-26 · Official public authority

    Supports
    Documents coordinated cloned-media infrastructure and distribution patterns attributed by German authorities to the Doppelgänger campaign.
    Does not establish
    Does not by itself establish unique reach, persuasion, or the independent contribution of generative AI.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  3. VIGINUM · 2024-02-12 · Official public authority

    Supports
    Documents a coordinated network of information portals and its observable infrastructure and content-distribution patterns.
    Does not establish
    Does not establish audience belief or behavior change; AI involvement varies by component and must not be assumed from coordination alone.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  4. Recorded Future / Insikt Group · 2024-05-09 · Independent specialist analysis

    Supports
    Documents an inauthentic media network using large language models to modify and publish political content at scale.
    Does not establish
    Attribution and infrastructure findings are analytical assessments; social amplification and persuasive effect were limited or not established.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.
  5. SRC-07-GRAPHIKA-WOLF-NEWSDeepfake It Till You Make It

    Graphika · 2023-02-07 · Independent specialist analysis

    Supports
    Documents limited use of AI-generated fictitious news presenters in content promoted by a pro-China influence operation.
    Does not establish
    Does not establish substantial reach or persuasive effect and should not be generalized to all synthetic presenter use.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  6. Federal Communications Commission · 2024-09-30 · Official public authority

    Supports
    Confirms an AI-generated voice-cloning robocall campaign targeting New Hampshire primary voters and the FCC enforcement response.
    Does not establish
    Does not establish how many recipients believed the message or whether it changed turnout; it does not independently verify separate election-audio incidents in other countries.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.
  7. SRC-09-GOLDSTEIN-AI-PROPAGANDAHow persuasive is AI-generated propaganda?

    PNAS Nexus · 2024-02-20 · Primary research

    Supports
    Finds that AI-generated propaganda text can be persuasive in a controlled experiment and that human-machine curation can improve outputs.
    Does not establish
    A bounded experiment does not establish real-world campaign deployment, durable attitude change, behavior, or strategic effect.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  8. European External Action Service · 2024-01-23 · Official international public authority

    Supports
    Records reported AI-generated audio incidents in election contexts including Slovakia and places AI use within a bounded sample of 750 investigated FIMI incidents, with fewer than twenty involving AI and most AI use focused on content creation or legitimacy.
    Does not establish
    Does not by itself identify every creator or sponsor, prove that the Slovak audio changed votes or turnout, establish unique reach or exposure, or show that AI transformed the overall FIMI threat landscape.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Citation and source-scope review completed for WIP.51 on 2026-07-27.

  9. Harvard Kennedy School Misinformation Review · 2024-08-22 · Independent academic analysis

    Supports
    Documents the reported fake audio surfacing two days before the 2023 Slovak election and analyzes timing, encrypted-channel circulation, political amplification, institutional distrust, and competing explanations for the result.
    Does not establish
    Does not identify the audio creator or sponsor, provide a counterfactual election estimate, or establish that the audio swung the election; it explicitly treats that conclusion as reductive.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Citation and source-scope review completed for WIP.51 on 2026-07-27.
Selected works identified by the owner-supplied report
  • Goldstein et al., research comparing large-language-model propaganda with human-authored covert propaganda.
  • Graphika reporting on Spamouflage and synthetic “Wolf News” presenters.
  • Wack et al., analysis of generative rewriting and publication volume in a state-backed outlet.
  • Chesney and Citron, Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security.

Exact source preservation and editorial currentness review do not constitute specialist certification, adjudication, legal advice, clinical review, or proof that every owner-report citation is current. Corrections remain open.

Evidence methodReach versus effectCorrectionsDefensive incident template

Page complete AI-Generated Propaganda Page label: CONTEMPORARY / ONGOING CLAIM — NOT SETTLED HISTORY