Category 01 of 12 · Tool level
Human planners retain strategic control while AI assists with research, translation, production, audience analysis, simulation, or measurement inside an established influence workflow. Evidence trail for AIP-01-0001: 5 selected sources, 3 with a bounded automated source check.
Definition and boundary
Human-led psychological or strategic-influence activity in which AI supports bounded stages such as information synthesis, translation, content drafting, simulation, distribution support, or evaluation. Evidence trail for AIP-01-0002: 5 selected sources, 3 with a bounded automated source check.
Fully autonomous systems choosing political objectives without human direction; transparent public communication is not automatically a psychological operation merely because AI assisted its production. Evidence trail for AIP-01-0003: 5 selected sources, 3 with a bounded automated source check.
Strategic and human significance
The immediate change is organizational: tasks that once required large multilingual teams can be performed faster and at lower cost. Evidence trail for AIP-01-0005: 5 selected sources, 3 with a bounded automated source check.
The same systems can support lawful public communication or deceptive operations, so intent, attribution, consent, and oversight remain decisive. Evidence trail for AIP-01-0006: 5 selected sources, 3 with a bounded automated source check.
AI can reduce production friction and accelerate feedback loops while human operators retain strategic intent and accountability. Evidence trail for AIP-01-0004: 5 selected sources, 3 with a bounded automated source check.
Change from pre-AI practice
Evidence maturity
The report identifies confirmed use of generative AI by state-aligned and influence-for-hire operators, while noting that measured engagement and durable persuasion remain separate questions.
Reports from platforms and public authorities describe AI used for drafting, translation, persona support, research, and technical assistance in covert influence activity. Evidence trail for AIP-01-0012: 5 selected sources, 3 with a bounded automated source check.
Controlled studies show that conversational models can be persuasive under bounded conditions, especially when given target information. Evidence trail for AIP-01-0013: 5 selected sources, 3 with a bounded automated source check.
High output, reach, or engagement does not establish durable belief or behavior change. Evidence trail for AIP-01-0014: 5 selected sources, 3 with a bounded automated source check.
Conflicting findings and scope boundaries
These are not errors to hide. They reflect different study designs, measures, time horizons, and operational contexts.
TENSION-PERSUASION-01
Evidence position A: A preregistered short conversational study found a stronger agreement shift when GPT-4 had basic participant information.
Evidence position B: Other report material questions whether psychographic microtargeting reliably adds persuasive value over high-quality generic messages in broader settings.
Publication rule: State the design, sample, treatment, and measured outcome. Do not generalize short-term agreement shifts into durable behavior change or population control.
TENSION-REACH-01
Evidence position A: Generative AI can reduce production friction, expand language coverage, and support high-volume content variation.
Evidence position B: OpenAI's 2024 disruption report found no meaningful authentic breakout for the five observed operations.
Publication rule: Report production, distribution, authentic reach, engagement, persuasion, and behavior as separate measures.
Conceptual mechanisms
Machine-assisted audience and narrative analysis. Evidence trail for AIP-01-0015: 5 selected sources, 3 with a bounded automated source check.
Human-reviewed generation and localization. Evidence trail for AIP-01-0016: 5 selected sources, 3 with a bounded automated source check.
Synthetic pretesting and scenario comparison. Evidence trail for AIP-01-0017: 5 selected sources, 3 with a bounded automated source check.
Automated scheduling and measurement. Evidence trail for AIP-01-0018: 5 selected sources, 3 with a bounded automated source check.
Human approval and deployment remain the defining control point. Evidence trail for AIP-01-0019: 5 selected sources, 3 with a bounded automated source check.
These descriptions explain capability and risk. They intentionally omit procedures, targeting criteria, scripts, and evasion methods.
Evidence and examples
Evidence trail for AIP-01-0020: 1 selected sources, 1 with a bounded automated source check.
Evidence trail for AIP-01-0021: 1 selected sources, 1 with a bounded automated source check.
Case studies show documented events or bounded experiments. They do not establish prevalence, general causation, or guaranteed persuasive effect.
Risk and failure analysis
Detection and defense
Caveat: Coordination can also reflect legitimate campaigns or shared news events. Evidence trail for AIP-01-0030: 5 selected sources, 3 with a bounded automated source check.
Caveat: This is strong evidence of automation error but not proof of a specific sponsor. Evidence trail for AIP-01-0031: 5 selected sources, 3 with a bounded automated source check.
Caveat: Professional communications teams also adapt quickly. Evidence trail for AIP-01-0032: 5 selected sources, 3 with a bounded automated source check.
Governance and safeguards
Open questions
Sources and evidence boundary
Primary synthesis: AI-Assisted Psychological Operations Report. The complete report is retained in a non-public provenance directory with SHA-256 1eec5ce4b5e99479617b506bab22986821a10c3b4c8d6d6daba32b64d26623d1.
Thirty high-impact references received bounded automated retrieval, official corroboration, or stronger-source substitution. All 91 selected references are used by the 501-claim citation graph, but the full report corpus and human editorial acceptance remain unverified. Source type labels are editorial classifications, not quality scores.
The source describes a preregistered, bounded multi-round persuasion experiment. The 81.7 percent result applies to odds of increased agreement in the stated experimental condition, not population-level mind control.
Limits: Controlled, short-horizon setting with 820 unique participants. The result does not establish durable real-world behavior change or operational effectiveness.
The German Foreign Office report documents technical indicators associated with the Doppelgänger campaign and supports a bounded claim that the operation used coordinated deceptive infrastructure. The source supports campaign-attribution and infrastructure claims within the report's stated scope; it does not establish durable persuasion effects.
Limits: A government technical report reflects the collecting institution's visibility and attribution methodology. Operational reach and content volume are not evidence of attitude or behavior change.
OpenAI reported disrupting five covert influence operations that used its models for content, translation, research, and code assistance. The same report stated that those operations did not appear to gain meaningful authentic reach or engagement from the service.
Limits: Provider-authored threat reporting covers activity visible to that provider. No claim is made that all campaign activity or downstream effects were observed.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.