Donald Trump’s use of AI-generated media, particularly content that fabricates or manipulates the speech and actions of political opponents, constitutes a form of democratic erosion by proxy: it does not directly dismantle democratic institutions, but it degrades the norms of mutual toleration and pluralism that make those institutions function.

Nancy Bermeo’s framework of democratic backsliding is especially useful here. In “On Democratic Backsliding,” Bermeo emphasizes that modern democratic erosion rarely occurs through overt coups; instead, it proceeds through subtler mechanisms that hollow out democracy while preserving its formal shell. Trump’s AI usage fits squarely within this pattern. Rather than suspending elections or dissolving courts, AI-generated depictions of figures like the Obamas fabricate political reality itself, undermining citizens’ ability to distinguish legitimate opposition from constructed enemies.

This tactic does not eliminate democratic choice outright, but it distorts the informational environment in which choice is exercised. As Bermeo argues, democracy depends not only on rules but on shared understandings of political legitimacy. When synthetic media falsely portrays opponents as criminal, treasonous, or conspiratorial, it weakens the normative commitment to opposition as a legitimate component of democracy.

This dynamic aligns closely with Jan-Werner Müller’s account of populism in What Is Populism?. Müller defines populism not by anti-elitism alone, but by anti-pluralism: the claim that only one group represents the “real people,” while opponents are illegitimate or corrupt by nature. AI-generated media amplifies this logic by allowing populist leaders to fabricate “evidence” that reinforces moral exclusion. When Trump circulates AI content depicting political rivals saying or doing things they never did, he is symbolically expelling them from the democratic community.

Importantly, this is not just persuasion, it is delegitimization. Müller warns that populists undermine democracy by rejecting the idea that political opponents can lose elections legitimately or govern lawfully. AI accelerates this process by collapsing the boundary between critique and invention. The danger lies not in voters being misled once, but in the cumulative erosion of the idea that democratic competition occurs among equals.

Steven Levitsky and Daniel Ziblatt’s analysis in How Democracies Die further clarifies why this matters. They argue that democracies depend on informal norms particularly mutual toleration and institutional forbearance. Trump’s AI usage violates mutual toleration by portraying rivals as existential threats rather than legitimate competitors. Even when institutions formally remain intact, these norm violations weaken democracy’s foundations.

Crucially, AI allows these violations to scale. Whereas earlier norm breaking relied on speeches or social media posts, synthetic media offers a level of realism and virality that blurs the line between fiction and documentation. This makes norm erosion more resilient to fact checking or institutional correction. Courts can rule, and elections can proceed, but public belief in their legitimacy erodes underneath.
Ozan Varol’s concept of “stealth authoritarianism” also applies. Varol describes how leaders use legal or seemingly benign tools to entrench power while maintaining democratic appearances. AI fits this model precisely: it is not illegal per se, nor is it inherently authoritarian. Yet when deployed strategically by incumbents or dominant political figures, it becomes a means of manipulating democratic competition without formally violating democratic rules.

Skeptics might argue that voters are capable of identifying AI-generated content and discounting it accordingly. This is the core empirical question that makes the argument falsifiable. If AI exposure does not measurably affect trust in democratic institutions, opposition legitimacy, or willingness to accept electoral outcomes, then its role in democratic erosion is limited. But if it does, as existing research on misinformation suggests, then AI represents not a novel threat, but a powerful extension of familiar ones.
Trump’s use of AI should not be understood as a technological curiosity or mere campaign tactic. Through the lens of democratic erosion theory, it is best seen as a norm-destroying tool that accelerates populist anti-pluralism while preserving democratic form. Democracy may survive the appearance of AI politics but it must be weary when it’s time to survive the reality it constructs.