How AI Slop Is Powering Influence Campaigns
Key Findings
- When generative AI tools became widely accessible, the research community anticipated that AI integration would lead to higher-quality and more sophisticated AI-enabled influence operations (IOs) that could deceive and mislead audiences, especially around elections.
- Research on AI-enabled IOs, including the case studies documented in this report, indicates that the production of these campaigns has increased in speed and scale. However, while AI tools helped to amplify AI-generated text and footage across social media platforms, the quality and sophistication of this content remain low.
- The case studies here discuss well-documented IOs, including CopyCop, Doppelgänger, Spamouflage, Falsos Amigos, Operation Overload, Operation Undercut, and a pro-India celebrity impersonation campaign. Taken together, they demonstrate that manipulative state and non-state actors are actively seeking to use AI tools to deceive and mislead target audiences by amplifying polarizing or sensational content. Their output is often low-quality, yet scalable content that poses as fake and authentic individuals and institutions.
- Using the Actor, Behavior, Content (ABC) model, we analyzed IOs attributed to state and non-state actors to identify their most common characteristics. We examined the Tactics, Techniques, and Procedures (TTPs) deployed in these operations for indicators of increased use or reliance on AI tools to conduct and scale their operations.
- The sponsors behind publicly documented IOs have delegated core functions such as content creation and persona generation to AI tooling. These campaigns then saturate the information ecosystem with low-quality content but highly scalable campaigns that blur the lines between professional and amateur actors.
- The AI-enabled IOs documented by Graphika and others show that, based on engagement metrics, their ability to reach organic audiences remains limited.