In September, 2025, thousands marched through London under the banner of the far-right activist Tommy Robinson. There were English flags, nationalist slogans, and the familiar theatre of political grievance. But the most revealing aspect of the gathering was not the spectacle itself. It was the extraordinary range of beliefs that appeared to unite people who otherwise had little in common.

The journalist Peter Geoghegan, writing in Democracy for Sale, described encountering similar attitudes at Reform UK's conference in Birmingham. Climate change was supposedly a scam. Covid vaccines had allegedly caused cancer in the Royal Family. The Fabian Society secretly controlled Britain. These conspiracies differed in almost every detail, yet shared one conviction: established institutions were lying, and only outsiders understood the truth.

In Geoghegan's neighbourhood, St George's crosses appeared in red paint on doors, predominantly those of immigrant families. The distance between political fantasy and everyday intimidation had become uncomfortably small. Something larger than British nationalism was unfolding. The internet had not invented extremism, but it had transformed the conditions under which extremism could circulate, recruit, and acquire legitimacy.

The Business of Distrust

Eliot Higgins, founder of the investigative organization Bellingcat, describes this environment as disordered discourse. Its defining characteristic is not simply misinformation, but the systematic erosion of confidence in how knowledge is established. When every journalist is supposedly compromised, every scientist corrupt, and every contradiction evidence of conspiracy, reality becomes whatever one's political community agrees to believe.

The philosopher Hannah Arendt recognized a related danger in her examination of totalitarianism: the destruction of distinctions between fact and fiction can leave populations vulnerable to manipulation. Contemporary democracies are not necessarily becoming totalitarian states, but the weakening of shared standards of evidence presents a genuine democratic problem.

Much of this distrust has legitimate historical origins. The Iraq War, the financial crisis, political corruption, and decades of broken promises damaged institutional credibility long before social media platforms perfected their recommendation systems. Algorithms did not manufacture public disappointment. They entered a political landscape already saturated with it.

Their contribution was to make that disappointment extraordinarily profitable.

Social media companies operate systems designed to predict and sustain attention. Anger, fear, and sensationalism can flourish within those systems, particularly when political entrepreneurs learn how to exploit their incentives. A conspiracy need not persuade everyone. It merely needs to reach enough receptive people, repeatedly, until suspicion becomes a shared identity.

The Invisible Editors of Democracy

For much of modern history, newspapers and broadcasters exercised considerable influence over public debate. Their editorial judgments were often biased, sometimes dishonest, and frequently shaped by powerful interests. But the institutions making those decisions were identifiable.

Today, much of what citizens encounter is selected by algorithms operating behind proprietary systems. Two neighbours can inhabit the same city, vote in the same election, and experience profoundly different political realities.

The scholar Cass Sunstein has warned that fragmented information environments can weaken democratic deliberation. Political scientist Nardine Alnemr, writing in Democratization, identifies three related dangers: exclusion from competing perspectives, the amplification of unreliable information, and manipulation through automated accounts and repetition.

The evidence does not establish that recommendation algorithms inevitably radicalize users. People actively choose political communities, and offline circumstances remain crucial. Nevertheless, the power to determine which information becomes visible has migrated toward institutions that citizens neither elect nor meaningfully oversee.

This is a constitutional problem disguised as a technological convenience.

Democracy rests on more than counting votes. It requires citizens capable of evaluating competing claims, scrutinizing authority, and participating in a public conversation. When the architecture of that conversation belongs to a handful of corporations, democratic sovereignty becomes incomplete.

Artificial intelligence intensifies the difficulty. Synthetic videos, cloned voices, and automated political messages can make deception cheaper and more convincing. Equally troubling is the possibility that authentic evidence can be dismissed as artificial whenever it becomes inconvenient.

The result is not merely a society that believes too many lies. It is a society increasingly uncertain whether truth can be established at all.

Who Governs the Algorithm?

The obvious remedies include platform transparency, independent audits, digital literacy, and stronger protections against coordinated manipulation. These are necessary, but none can substitute for rebuilding public trust through accountable government and credible journalism.

There are more imaginative possibilities. Taiwan's collaborative fact-checking initiative Cofacts demonstrates how citizens can participate directly in evaluating suspicious information. Rather than treating the public as passive consumers awaiting correction, such approaches make verification a shared civic practice.

Democratic institutions must also confront the economic and social grievances that conspiracy entrepreneurs exploit. A society experiencing declining security, widening inequality, and institutional indifference cannot expect fact-checking alone to restore confidence.

The deeper challenge is political. Governments regulate elections, campaign financing, and broadcasting because the organization of public communication matters to democratic legitimacy. Algorithmic systems now exercise comparable influence over political visibility, yet their governing decisions remain largely private.

We have allowed the infrastructure of public opinion to become a commercial product while continuing to imagine democracy as something that happens primarily inside parliament and at polling stations.

The ballot box remains indispensable. But elections cannot fully protect democratic self-government when the information shaping political judgment is governed elsewhere.

We still elect the people who govern our countries. We have never elected the systems increasingly governing what we believe about them.