Think about how much information you scroll past in a single day. News alerts, TikTok videos, tweets, headlines, captions, and posts from people you have never met. Now imagine that a significant portion of that content was not created by humans at all, but by machines designed to influence how you think. That is not a distant dystopian fantasy. It is happening right now. In 2025, research from Graphika revealed that online propaganda campaigns are increasingly built on artificial intelligence for the most basic functions of influence operations: generating content, running fake accounts, and creating inauthentic influencer personas. The report described fake social media profiles, videos featuring synthetic news reporters, and entire fake news websites filled with AI-generated articles. What used to require a room full of writers, editors, actors, and social media managers can now be done by an algorithm working around the clock. This changes everything. It changes how quickly disinformation can spread, how cheaply it can be produced, and how difficult it is to tell who is actually behind a message. The familiar face on a video might be a digital ghost, assembled from reused footage and recycled voices. The blog post that seems to be written by a concerned citizen might be the output of a language model. The comment section that appears to show a lively debate could be full of bots talking to one another. None of this means that all AI-generated content is bad. Many useful tools use the same technology. But the same systems that can draft essays and generate realistic images can also be weaponized to manipulate public opinion. And in Southeast Asia, this weaponization is already underway.
One of the most striking findings from recent investigations is that hostile actors do not always need to invent new controversies. They can simply take existing social, political, or economic fault lines and pour fuel on them. Every country has tensions: anxiety about jobs, frustration over housing costs, worry about foreign investment, fear of losing national identity. These are the raw materials of influence operations. With artificial intelligence, malicious actors can generate enormous volumes of content along these narratives, often referred to as “AI slop.” The term may sound silly, but the impact is serious. In Singapore, for example, journalists uncovered AI-generated female personas on TikTok that used reused voices and recycled scripts. The videos framed Singapore as economically reliant on China, a narrative carefully chosen to provoke distress and resentment. What made the content especially dangerous was that it blended genuine news with false and misleading claims about Singapore’s economy. This is not an easily recognizable fake; it is a distorted mirror. Some of the facts are real, which gives the fabricated parts an uncomfortable credibility. Investigators found signs of coordinated behavior, such as multiple videos published within a short period of time, suggesting a centralized operation rather than independent users. Similar narratives appeared on YouTube. And the campaign did not stop at Singapore’s borders. Videos targeting audiences in Malaysia were also discovered. This pattern is not random. It reflects a deliberate effort to saturate the regional information environment with content designed to exploit vulnerabilities and deepen divisions.
What is the real goal of such campaigns? Many people assume that the goal is to make a target audience believe a specific lie. But that assumption may be too simple. The objective is often far more subtle and far more corrosive: to shape the broader information environment in which citizens form their opinions. When people see the same narrative repeated across multiple videos, posts, and articles, that repetition creates familiarity. And familiarity, as psychologists have long known, breeds acceptance. You might not remember where you first saw a claim, but if you see it enough times, it starts to feel true. Over time, repeated exposure can turn a misleading narrative into something that seems like common sense. But there is another effect as well: uncertainty and distrust. When citizens are flooded with content whose origins they cannot verify, they begin to question everything. Is this source legitimate? Is this video authentic? Is this journalist a real person or a synthetic creation? This erosion of trust is exactly what the operators want. A society that cannot agree on basic facts cannot effectively solve its problems. It becomes paralyzed, suspicious, and easy to manipulate. The shared information environment that makes democracy possible—the space of news, social media, and public conversation—becomes polluted. It is like a river full of sediment. You cannot see the bottom. You cannot tell what is safe to drink. Eventually, you stop drinking altogether. That is the true danger of AI slop. It is not just that a falsehood is repeated. It is that the public loses faith in the very idea of truth. In this way, AI disinformation is not about winning an argument. It is about making argument impossible.
Tech and social media companies are beginning to recognize the threat and are trying to respond. YouTube and LinkedIn, for example, have been removing low-quality spam and updating their community guidelines to limit the spread of AI slop. Meta and TikTok have introduced automated labels for synthetic media, so users can see when content has been generated or altered by artificial intelligence. More companies are incorporating SynthID, a standard of detection that allows individuals to check whether an image was produced by AI. These are encouraging steps, but they are not enough. Enforcement across platforms and products remains varied and inconsistent. A label might appear on one video but not on another. A spam filter might work effectively on YouTube but fail elsewhere. Meanwhile, algorithmic feeds continue to reward AI-generated posts with high engagement rates. Why? Because AI slop is often designed to provoke strong emotional reactions, and engagement is the currency of the digital economy. A platform may claim it wants to limit misleading content, but its own recommendation systems continue to amplify it. This is a painful contradiction at the heart of the online world. Platforms are caught between serving their users and serving their business models. As long as AI-generated outrage drives clicks, there will be an incentive to let some of it through. The result is a patchwork response. Some companies are doing serious, thoughtful work. Others are doing the bare minimum. And all of them are struggling to keep pace with a technology that evolves faster than any set of guidelines can.
In this difficult landscape, a layered approach to resilience is not just useful—it is essential. No single solution, whether technological, regulatory, or educational, can fully solve the problem. At the government and institutional level, monitoring of recurring narratives and coordinated networks can be enhanced. Intelligence agencies, policy makers, and civil society organizations can track patterns of behavior, identify campaigns, and expose the networks behind them. But governments cannot do this alone. Continued media and information literacy is perhaps the most important defense we have. People need to understand what artificial intelligence can do, how it can be misused, and how to question the content they see online. This should be a top priority for governments, educators, and civil society alike. It means teaching students to verify sources, to notice suspicious patterns, and to ask who benefits from a particular message. It also means reaching adults who did not grow up with these tools and may not be aware of how realistic fake content has become. The challenge goes beyond detecting AI slop. It is about enhancing a society’s resilience to the narratives and influence operations that the content carries. In other words, the goal is to build a population that can withstand the flood. A population that can separate signal from noise. A population that does not panic at every misleading headline and does not retreat into cynical indifference. This kind of resilience is not a one-time program. It is an ongoing investment in the public good, and it requires the participation of everyone.
We are at a crossroads. The same artificial intelligence that can compose poetry, diagnose diseases, and help scientists make breakthroughs can also be used to manipulate publics and undermine democratic institutions. The technology itself is not evil; it is a tool. But like any tool, it reflects the intentions of those who wield it. Hostile actors have already shown that they will use AI to exploit our differences. They will fish in troubled waters, using our own anxieties and grievances against us. The answer cannot be to abandon technology, and it cannot be to live in a state of constant suspicion. The answer is to approach the digital world with open eyes. We need better safeguards from tech companies, stronger oversight from governments, and a more informed and engaged public. We need to hold onto the human connections and critical thinking that algorithms cannot replicate. And we need to recognize that defending the information environment is just as important as defending our physical borders, because in the twenty-first century the two are deeply intertwined. There is no finish line here. AI will keep evolving, and so will the tactics of those who misuse it. But if we stay awake, stay curious, and stay committed to the truth, we have a chance. Not a perfect chance, but a real one. The future of public conversation depends on the choices we make now—as individuals, as institutions, and as societies. The digital world is ours to shape. We should shape it wisely, before someone shapes it for us.