It spreads through our information environment like a virus, and like a virus, it is far cheaper to prevent than to cure.
When societies finally accepted that smoking was a public health problem, something important shifted. The question stopped being only whether one person chose to light a cigarette and became a question about all of us: the air in shared spaces, the burden on our health systems, the advertising that made the habit look normal. We did not reduce smoking by scolding individual smokers. We reduced it, to the extent we have, by treating it as a societal risk. That meant prevention, education, regulation of the industry, and a slow, deliberate change in what society was willing to tolerate.
I want to make the case that ambient radicalisation deserves the same lens.
In my recent case study for the Institute of Communication Studies, I described how the older models of radicalisation are, in effect, targeted. Extremist movements focus on specific, often already-disenfranchised individuals, draw them in through recruiters and networks, and move them along an identifiable pathway toward a rigid worldview. There is a message, a messenger, and a movement. That model still exists, but it no longer captures most of what is happening in online spaces, especially when Foreign Interference comes into the mix.
What has changed is the goal. The aim is no longer only to push one individual toward a violent act. It is to build capacity: to erode trust in institutions across a whole population, narrow the space for disagreement, and shift the point at which violence starts to look acceptable, or eventually normal. Rather than hunting for the right recruit, actors push polarising content into the open and let it circulate, knowing it will land with whoever has the grievances to make it resonate. Nobody needs to be recruited into anything. The target is not a narrow at-risk group. The target is society itself.
This is where the public health lens earns its place, because we already know a great deal about how this particular virus transmits.
Let’s start with the environment. Social media rewards emotional and polarising language, so the most divisive material travels furthest almost by default. Vosoughi, Roy and Aral’s 2018 analysis in Science of some 126,000 stories shared on Twitter over a decade found that false news reached more people and spread faster than the truth in every category examined, and was around 70 percent more likely to be reshared. Emotion is the mechanism. Studying over half a million posts on gun control, climate change and same-sex marriage, Brady and colleagues found each additional moral-emotional word increased a message’s diffusion by roughly 20 percent — later replications put it nearer 13 to 17 percent, but consistently positive. Crucially, the boost operated mostly within ideological groups rather than across them. Emotional content does not just travel further. It travels further into the echo chamber.
Now add the actors. The European External Action Service documented 540 incidents of Foreign Information Manipulation and Interference during 2025 alone, involving some 10,500 channels and websites. But the method matters more than the volume. These campaigns rarely invent grievances; the EEAS describes them as targeting vulnerable groups and amplifying prejudices that already exist. Its threat reports make the link to ambient radicalisation explicit: individual incidents are often limited in effect, yet the constant, daily accumulation of manipulative behaviour has a long-term corrosive effect on societies and on trust in democracy. Manipulation is less an injection of foreign ideas than an accelerant poured on tensions already burning.
The damage is measurable. The OECD’s 2023 survey of nearly 60,000 people found more reporting low or no trust in their national government (44 percent) than high trust (39 percent), a decline since 2021 that the OECD links directly to an information environment saturated with polarising content. The Open Society Barometer, meanwhile, found 35 percent of 18-to-35-year-olds saying a strong leader who dispenses with elections and parliaments was a good way to run a country, against 26 percent of those over 56. The generation most immersed in this environment is the least convinced that democracy can deliver.
Occasionally the infection becomes acute. The attempted storming of the Reichstag in August 2020 was carried out by several hundred people peeled off a 38,000-strong protest against pandemic measures — an eclectic mix of anti-vaccine activists, Reichsbürger, QAnon adherents and the far right, gathered under narratives of a “corona dictatorship.” Four months later came the US Capitol; two years after that, Brasília. Comparative research now reads these as one transnational pattern, in which conspiracy narratives work as bridging stories connecting otherwise unrelated milieus. Most tellingly, analysis of both the Washington and the Brasília attacks concluded that unaffiliated participants outnumbered members of organised groups: exposure to circulating disinformation motivated people more effectively than belonging to any movement with a leader and a logo.
These, to my mind, are examples of ambient radicalisation. Nobody recruited most of those people. The environment did the work, and if that is the diagnosis, we should focus on an honest treatment plan, and not fall into two specific traps.
The first is to blame the patient. Hunting for the individual who has been “radicalised” and treating them as the whole problem is like tackling lung cancer while ignoring the tobacco industry. The structural driver is an attention economy in which bad behaviour is systematically rewarded. Inflammatory politicians, dehumanising content and deliberate provocation are amplified by the same algorithms that decide what the rest of us see. Any serious response has to address that incentive structure, not just its victims.
The second is to imagine regulation alone will cure us. It will not, but it is still part of the treatment. The EU’s Digital Services Act can compel platforms to take systemic risks seriously and constrain the infrastructure through which manipulation spreads. That is real harm reduction. What it cannot do by itself is rebuild trust or reduce polarisation, any more than banning cigarette advertising cured anyone already ill or addicted to the product.
One of the tools we have is, fittingly, borrowed straight from medicine. “Prebunking” — exposing people in advance to weakened forms of a manipulation technique so they recognise it later — derives from inoculation theory, and the name is no coincidence. In a YouTube field study with more than 22,000 users, Roozenbeek, van der Linden and colleagues found that ninety-second videos on techniques like emotionally manipulative language, scapegoating and false dichotomies measurably improved people’s ability to spot manipulation. Researchers now speak, only half-metaphorically, of psychological herd immunity.
It works best as one layer among several. Digital and media literacy teach people how information is produced, funded and distributed. Critical thinking teaches them to interrogate a claim. But the most underrated layer is emotional literacy: the capacity to notice what a piece of content is doing to you, and to pause when something makes you feel fury or existential threat. Manipulation rarely works by convincing us of an argument, it works by provoking a feeling we then rationalise. Recognising that feeling as it arrives, and asking who benefits from my outrage, is the closest thing we have to an immune response.
The frame carries one warning, and it matters especially in this region. The Council of Europe has cautioned that online-harm regulation can itself be abused where oversight is weak and polarisation high. A treatment that damages the patient is no treatment. Norm-setting around manipulation must protect democratic debate rather than police dissent, target manipulative behaviour and not oppositional viewpoints.
For North Macedonia and the wider Western Balkans this is not abstract. The region combines historically charged fault lines, fragile information environments, and a long, frustrating relationship with European integration. Real disappointment is precisely what manipulation attaches itself to most easily; vulnerability is greatest when the grievance is real.
We came to understand that smoking was not only bad for the smoker but costly for society as a whole. Ambient radicalisation works the same way: a risk to the individual, and a mounting, measurable cost to all of us. We are already paying it. The question is whether we keep treating symptoms one crisis at a time, or start investing, seriously, in immunity.
Jordy Nijenhuis is Director of Dare to be Grey, a Dutch foundation working on the prevention of Polarisation, Disinformation and Hate. This article draws on his case study “Ambient Radicalisation: How FIMI Exploits Polarisation,” written for the Institute for Communication Studies.
