AI Propaganda Isn't Just a Content Problem. It's an Emotional Infrastructure Problem
You've probably seen it without recognising it. An AI-generated rapper dropping bars about immigration and cultural decline. A Lego-style animation mocking world leaders. A meme that feels weirdly funny and unsettling at the same time. This is slopaganda — AI-generated propaganda designed to manipulate beliefs and feelings at speed and scale — and it is spreading fast.
A new piece from the Media Diversity Institute features Professor Vian Bakir (Bangor University) on exactly this phenomenon. The MDI article traces how slopaganda is being used to target marginalised groups, simulate false consensus, and flood information environments with emotionally charged content that is cheap to produce, easy to share, and hard to counter.
But here's what the slopaganda debate often misses: the content itself is only the surface.
The Infrastructure Underneath
Slopaganda works not by presenting arguments people consciously accept, but by quietly embedding emotions and associations below the threshold of awareness. Repeated exposure to the same images, narratives, and emotional cues shapes how people feel about groups, institutions, and ideas — often without their knowledge or consent. The manipulation is not in any single piece of content. It is in the system that delivers it.
That system is emotional AI.
In their chapter "Guarding against Automated Empathy Attacks on Ontological Security" in the Routledge Handbook of the Influence Industry, Vian Bakir and Andrew McStay map out how emotional AI technologies — from social media sentiment profiling to biometric emotion recognition and, on the horizon, neurotech — can be weaponised to erode something far deeper than a person's opinion on a policy issue. They can erode ontological security: our foundational sense of self, identity, trust, and place in the world.
This is not hypothetical. The groundwork is already laid.
From Feeds to Bodies to Brains
Social media platforms have spent years perfecting the art of emotional engagement — using likes, reactions, and algorithmic feedback loops to understand which content produces which emotional responses in which people. That data is already being used to microtarget political messages designed not to inform, but to provoke.
Now layer in biometric emotion recognition: systems that read facial expressions, vocal patterns, and physiological signals to infer how a person feels in real time. Then consider the trajectory towards neurotech — brain-computer interfaces and haptic wristbands that register intention and emotion before a person is consciously aware of either.
Bakir and McStay's chapter brings this trajectory into sharp focus through two scenarios set in a fictional country with lax data protection laws, showing how each layer of emotional AI — social media profiling, biometrics, neurotech — opens new avenues for malicious actors to target an individual's deepest anxieties, reinforce them, and steer behaviour at moments of civic importance such as elections and referenda.
The goal is not to win an argument. It is to reshape how reality feels.
The Social AI Dimension
The concern extends to social AI — the conversational agents, companion bots, GenAI influencers, and AI-driven social platforms increasingly woven into daily life. Where slopaganda broadcasts emotional manipulation at scale, social AI enables it one-to-one. A chatbot that learns your anxieties, mirrors your emotional state, and personalises its responses accordingly is, in effect, an intimate emotional profiling system. In the wrong hands — or simply under the wrong incentive structures — that intimacy becomes a vector. The same logic that makes a companion AI feel supportive can make a politically motivated one feel trustworthy, even as it quietly steers beliefs and behaviour.
Then there are GenAI influencers — synthetic personas with fabricated life stories, curated aesthetics, and algorithmically optimised emotional appeal — who build parasocial relationships with real audiences at scale, shaping opinions and purchasing decisions without ever disclosing their non-existence. As the MDI piece notes, AI influencers are already racking up devoted followings, with lonely or emotionally vulnerable audiences particularly at risk.
The emotional attunement that makes these systems feel real is precisely what makes them dangerous when deployed without adequate oversight.
Regulation Is Lagging — Badly
Both the MDI piece and the Bakir and McStay chapter point to the same structural problem: regulation is still oriented towards individual pieces of content, while the industry is already moving into emotions, bodies, and brains.
The EU AI Act and the Digital Services Act are steps in the right direction — requiring transparency around deepfakes, placing duties on platforms around systemic disinformation risks. But as Vian Bakir notes in the MDI piece, current policy is not keeping pace with AI-generated mechanisms for manufacturing false consensus at scale. Deepfake labels get stripped as content circulates. Emotional targeting operates invisibly. And the architecture of emotional profiling continues to expand.
Effective governance will need to be as systemic as the threat — addressing not just what content says, but how emotional infrastructure is built, deployed, and monetised.
Why This Matters to Us
At the Emotional AI Lab, understanding the governance of emotional AI technologies is central to everything we do. Our research tracks how these systems work, who they affect, and what safeguards are needed to protect civic life.
Slopaganda is one front of that concern. But it is part of a much larger picture — one in which the technologies designed to understand human emotion are increasingly being turned towards manipulating it.
The visible propaganda is alarming enough. It is the invisible infrastructure we most urgently need to understand.
Read the Media Diversity Institute piece on slopaganda → https://www.media-diversity.org/slopaganda-turns-stereotypes-into-content/
Read the Bakir & McStay chapter → https://www.routledge.com/Routledge-Handbook-of-the-Influence-Industry/Briant-Bakir/p/book/9781032188997