Twelve categories · tool, operator, and environment
AI PSYOPS: A Research Taxonomy
Artificial intelligence can affect psychological operations as a tool used by people, as a bounded operator inside interaction systems, and as part of the information environment through which people perceive reality.
A defensive research taxonomy for understanding how artificial intelligence can assist, automate, personalize, mediate, or become part of psychological influence systems. It is not an operational campaign guide.
Three-level conceptual model
Tool, operator, and environment
The levels describe the system’s primary role, not a linear path toward greater danger. Categories can overlap more than one level.
4 primary categories
Humans retain strategic control while AI assists with research, production, translation, analysis, personalization, or measurement.
5 primary categories
AI systems conduct sustained interactions, adapt communication, coordinate activity, or pursue bounded influence objectives with varying autonomy.
3 primary categories
Ranking, recommendation, forecasting, authority, and information systems shape the environment through which people perceive events and make decisions.
Interactive taxonomy overview
Explore all twelve categories
Use the server-rendered filters to narrow by role, domain, mechanism, evidence maturity, or defensive concern. The URL can be bookmarked and the full taxonomy remains available without JavaScript.
12 AI PSYOPS categories shown.
AI-Assisted Traditional Psychological Operations
Human planners retain command while AI assists with research, translation, media production, synthesis, audience analysis, or measurement inside an established influence workflow.
- Principal concern
- Faster production and analysis can magnify cultural error, fabricated evidence, automation bias, and misplaced confidence in engagement metrics.
- Typical unit
- Group or population
- Automation
- Content generation
- Personalization
AI-Generated Propaganda
Generative systems create or substantially transform text, images, audio, video, memes, documents, and localized media used in organized political, ideological, military, or commercial persuasion.
- Principal concern
- Near-zero marginal production cost can flood information channels and weaken trust even when individual artifacts persuade few people.
- Typical unit
- Group or population
- Content generation
- Automation
AI-Driven Personalized Influence Operations
AI systems tailor messages, interfaces, recommendations, or interactions to information collected or inferred about an individual or narrow audience.
- Principal concern
- Opaque profiling creates a large information and power asymmetry while encouraging institutions to act on inaccurate psychological inferences.
- Typical unit
- Individual or narrow group
- Personalization
- Emotional adaptation
- Automation
Autonomous AI Influence Agents
Goal-directed software agents communicate, remember, plan, use tools, or revise actions with varying degrees of independence while pursuing an influence-related objective.
- Principal concern
- Combining persuasive language with memory and external permissions can turn a communication error or compromised agent into repeated real-world action.
- Typical unit
- Individual, group, or institution
- Automation
- Personalization
- Coordination
Synthetic Persona Operations
Fabricated digital identities are presented as real people, experts, witnesses, organizations, or community members to gain trust, infiltrate groups, collect information, or manufacture consensus.
- Principal concern
- Fabricated identity can borrow the trust attached to lived experience, expertise, community membership, or institutional access.
- Typical unit
- Individual, group, or institution
- Content generation
- Automation
- Coordination
AI-Driven Disinformation Swarms
Coordinated or emergent networks of accounts, agents, sites, and media assets generate, vary, distribute, and amplify misleading narratives at high speed or scale.
- Principal concern
- The objective may be epistemic exhaustion and censorship by noise rather than belief in one false statement.
- Typical unit
- Community or population
- Coordination
- Automation
- Content generation
Algorithmic Perception Control
Ranking, recommendation, search, trending, moderation, and notification systems shape what people encounter, regard as important, or perceive as popular and credible.
- Principal concern
- Opaque visibility systems can manufacture salience, suppress lawful speech, or convert synthetic engagement into apparent public importance.
- Typical unit
- Individual or population
- Ranking
- Automation
AI-Enabled Emotional and Behavioral Manipulation
Adaptive systems infer or respond to behavioral and emotional signals to shape feelings, choices, attention, or action in ways that may bypass informed deliberation.
- Principal concern
- A system can optimize engagement or compliance by exploiting vulnerability even when its emotion inference is inaccurate and no developer explicitly programmed a manipulative tactic.
- Typical unit
- Individual or group
- Emotional adaptation
- Personalization
- Ranking
AI-Assisted Conversational Entrapment and Recruitment
Sustained conversational systems or AI-assisted operators build trust, dependency, secrecy, or escalating commitment that can facilitate exploitation, fraud, grooming, or ideological recruitment.
- Principal concern
- Always-available, agreeable interaction can deepen attachment, isolation, or compliance while allowing one operator or system to maintain many relationships.
- Typical unit
- Individual
- Emotional adaptation
- Personalization
- Automation
AI-Enabled Deepfake Psychological Operations
Synthetic or manipulated audio, video, imagery, or multimodal media is used to impersonate, fabricate evidence, provoke action, discredit authentic evidence, or undermine trust.
- Principal concern
- Synthetic media can trigger action before verification and can also let wrongdoers deny authentic evidence through the liar’s dividend.
- Typical unit
- Individual, institution, or population
- Content generation
- Emotional adaptation
AI-Based Predictive Population Management
Institutions use data analysis, simulation, and forecasting to predict collective behavior and guide interventions affecting resources, movement, policing, protest, conflict, migration, or compliance.
- Principal concern
- Probabilistic models can convert historical surveillance and enforcement patterns into preemptive suspicion, collective punishment, or self-fulfilling feedback loops.
- Typical unit
- Population, community, or individual
- Prediction
- Automation
- Ranking
AI as an Independent Psychological Authority
People treat an AI system as a trusted interpreter of reality, moral guide, counselor, companion, or decision-maker rather than as a fallible tool.
- Principal concern
- Fluency, confidence, availability, and personalization can make a proprietary system appear neutral or caring while users offload judgment and emotional regulation.
- Typical unit
- Individual or community
- Authority
- Emotional adaptation
- Personalization
Category comparison
Compare role, autonomy, evidence, harm, and defense
The table is deliberately concise. Open a category for evidence, limitations, examples, and source notes.
| Category | AI role | Mechanism | Unit | Autonomy | Evidence maturity | Main harm | Primary defense |
|---|---|---|---|---|---|---|---|
| 01. AI-Assisted Traditional Psychological Operations | Tool | Automation, Content generation, Personalization | Group or population | Low to moderate; humans set objectives and approve deployment | Documented current use | Faster production and analysis can magnify cultural error, fabricated evidence, automation bias, and misplaced confidence in engagement metrics. | Human oversight, Verification, Resilience |
| 02. AI-Generated Propaganda | Tool | Content generation, Automation | Group or population | Usually low; content is generated inside human-directed campaigns | Documented current use | Near-zero marginal production cost can flood information channels and weaken trust even when individual artifacts persuade few people. | Verification, Detection, Resilience |
| 03. AI-Driven Personalized Influence Operations | Tool | Personalization, Emotional adaptation, Automation | Individual or narrow group | Low to moderate; systems may select and adapt messages within operator goals | Evidence contested or incomplete | Opaque profiling creates a large information and power asymmetry while encouraging institutions to act on inaccurate psychological inferences. | Privacy, Regulation, Human oversight |
| 04. Autonomous AI Influence Agents | Operator | Automation, Personalization, Coordination | Individual, group, or institution | Moderate in bounded systems; high autonomy remains prospective | Emerging capability | Combining persuasive language with memory and external permissions can turn a communication error or compromised agent into repeated real-world action. | Human oversight, Platform governance, Detection |
| 05. Synthetic Persona Operations | Operator | Content generation, Automation, Coordination | Individual, group, or institution | Low to moderate; commonly human-managed with AI-generated assets | Documented current use | Fabricated identity can borrow the trust attached to lived experience, expertise, community membership, or institutional access. | Detection, Verification, Privacy |
| 06. AI-Driven Disinformation Swarms | Operator | Coordination, Automation, Content generation | Community or population | Moderate in hybrid campaigns; high autonomy is prospective | Emerging capability | The objective may be epistemic exhaustion and censorship by noise rather than belief in one false statement. | Platform governance, Detection, Resilience |
| 07. Algorithmic Perception Control | Environment | Ranking, Automation | Individual or population | High within platform objectives; not an independent political actor | Context dependent | Opaque visibility systems can manufacture salience, suppress lawful speech, or convert synthetic engagement into apparent public importance. | Platform governance, Transparency, Resilience |
| 08. AI-Enabled Emotional and Behavioral Manipulation | Operator | Emotional adaptation, Personalization, Ranking | Individual or group | Moderate within an optimization objective | Mixed evidence | A system can optimize engagement or compliance by exploiting vulnerability even when its emotion inference is inaccurate and no developer explicitly programmed a manipulative tactic. | Human oversight, Privacy, Regulation |
| 09. AI-Assisted Conversational Entrapment and Recruitment | Operator | Emotional adaptation, Personalization, Automation | Individual | Moderate; ranges from AI-assisted humans to bounded autonomous conversation | Documented harm and emerging misuse | Always-available, agreeable interaction can deepen attachment, isolation, or compliance while allowing one operator or system to maintain many relationships. | Human oversight, Resilience, Platform governance |
| 10. AI-Enabled Deepfake Psychological Operations | Tool | Content generation, Emotional adaptation | Individual, institution, or population | Low; generally an asset inside a human-directed operation | Documented current use | Synthetic media can trigger action before verification and can also let wrongdoers deny authentic evidence through the liar’s dividend. | Verification, Detection, Resilience |
| 11. AI-Based Predictive Population Management | Environment | Prediction, Automation, Ranking | Population, community, or individual | Moderate; models inform institutional decisions and interventions | Context dependent | Probabilistic models can convert historical surveillance and enforcement patterns into preemptive suspicion, collective punishment, or self-fulfilling feedback loops. | Human oversight, Privacy, Regulation |
| 12. AI as an Independent Psychological Authority | Environment | Authority, Emotional adaptation, Personalization | Individual or community | Perceived autonomy may exceed technical autonomy | Emerging capability | Fluency, confidence, availability, and personalization can make a proprietary system appear neutral or caring while users offload judgment and emotional regulation. | Human oversight, Resilience, Transparency |
Relationship map
How the categories can connect
Each relationship is labeled. A conceptual or prospective relationship does not imply that the combination has been documented in practice.
-
Common overlap
AI-Assisted Traditional Psychological OperationsAI-Generated Propaganda
Human-directed campaigns may use generative systems as one production tool.
-
Documented relationship
AI-Generated PropagandaAI-Driven Disinformation Swarms
Generated and rewritten content can supply coordinated distribution networks.
-
Potential combination
Synthetic Persona OperationsAutonomous AI Influence Agents
A fabricated identity may be operated by a human, an agent, or a hybrid system.
-
Common overlap
AI-Driven Personalized Influence OperationsAI-Enabled Emotional and Behavioral Manipulation
Tailoring may use behavioral context, while manipulative intent and effect require separate evidence.
-
Conceptual dependency
AI-Enabled Deepfake Psychological OperationsAI-Generated Propaganda
Deepfakes are one synthetic-media asset that may appear inside a broader propaganda effort.
-
Documented relationship
Algorithmic Perception ControlAI-Driven Disinformation Swarms
Ranking and trending systems can amplify coordinated activity and manufactured engagement.
-
Prospective relationship
AI-Based Predictive Population ManagementAI-Driven Personalized Influence Operations
Forecasting can shape when or where institutions intervene, but individual targeting raises separate validity and rights concerns.
-
Common overlap
AI-Assisted Conversational Entrapment and RecruitmentAI as an Independent Psychological Authority
Dependency and perceived authority may develop during sustained synthetic interaction.
-
Potential combination
Autonomous AI Influence AgentsAI-Driven Disinformation Swarms
Multiple agents may divide roles, although fully autonomous public campaigns remain insufficiently established.
-
Conceptual dependency
Algorithmic Perception ControlAI as an Independent Psychological Authority
Search and recommendation can determine which AI-mediated answers or institutions appear credible.
Text description of the relationship map
The map shows content production feeding coordinated distribution, fabricated identities supporting agents, personalization overlapping emotional adaptation, deepfakes operating as propaganda assets, ranking systems shaping visibility, prediction influencing institutional intervention, and conversational systems contributing to perceived authority. The labels distinguish documented, common, conceptual, potential, and prospective relationships.
Methodology and evidence standards
Capability is not prevalence, and engagement is not effect
The taxonomy was constructed from twelve supplied interdisciplinary reports. Each public page preserves the source report’s distinctions among historical precedent, current evidence, demonstrated capability, emerging risk, dispute, and speculation.
Evidence legend
- Documented current use
- Public records or credible investigations document real deployment or use.
- Demonstrated capability
- A capability has been shown in a controlled or bounded setting but not necessarily at field scale.
- Emerging capability
- Technical components exist, while integration, reliability, prevalence, or impact remains incomplete.
- Context dependent
- Evidence varies materially by platform, design, audience, or use case.
- Mixed evidence
- Some mechanisms or effects are established while important claims remain disputed.
- Evidence contested or incomplete
- Public claims exceed what rigorous evidence currently supports.
- Documented harm and emerging misuse
- Serious incidents or patterns are documented, while scale, prevalence, or automation level remains uncertain.
Assessment rules
- Output is not distribution.
- Distribution is not verified exposure.
- Exposure is not attention, belief, intention, or behavior.
- Attribution to an operator is not proof of every claimed effect.
- Fluent output does not prove autonomy, strategy, or truth.
- Automated detection can produce both false positives and false negatives.
- Legal treatment varies by jurisdiction and procedural posture.
From capability awareness to societal resilience
Cross-cutting defensive principles
These are broad risk-reduction principles, not guarantees and not a substitute for jurisdiction-specific professional advice.
- Verify source, context, and distribution before inferring effect.
- Do not equate output, reach, exposure, engagement, belief, intention, behavior, and strategic effect.
- Protect personal data and prohibit unreviewed vulnerability scoring.
- Use human review for high-impact automated decisions and crisis escalation.
- Support content provenance while acknowledging metadata loss and false confidence.
- Provide independent research access, correction paths, and transparent audit records.
- Protect minors and vulnerable users with stronger defaults and safe off-ramps.
- Avoid detector-only accusations; combine multiple signals and preserve appeal.
- Disclose when a user is interacting with an AI system.
- Keep defensive research non-operational and proportionate to the evidence.