X's Algorithm Pushes Users Toward Conservative Politics, Study Finds

A team of researchers from Italy, Switzerland, and France set out to answer a question that has haunted social media studies for years: Do algorithms actually change how people think about politics, or do they simply show us what we already believe? The answer, published in Nature, is more unsettling than either option.

The study tracked nearly 5,000 users of X during the summer of 2023, six months after Elon Musk acquired the platform formerly known as Twitter. Researchers randomly assigned participants to two groups: some used X's algorithmic feed, which the platform calls "For You," while others saw only the chronological timeline, the traditional reverse-chronological view of posts. For seven weeks, the researchers monitored what users saw, whom they followed, and how their political views shifted. They surveyed participants before and after the experiment, and used browser extensions to track their feed content and online behavior in real time.

The results were stark. Users assigned to the algorithmic feed engaged with the platform more frequently and shifted their political views noticeably to the right. They began following conservative political activists at higher rates. The algorithm itself was doing the heavy lifting: it surfaced more conservative content and activist posts while burying traditional news outlets. When researchers switched users from the algorithmic feed back to the chronological timeline, however, something unexpected happened. The rightward shift persisted. The users' political opinions and behavior barely budged, even after the algorithmic curation stopped. This suggests that exposure to the algorithm had already reshaped how users thought about politics in ways that outlasted the algorithm itself.

The effect was not uniform across all users. Those who identified as progressive or liberal experienced a relatively modest pull to the right. But users who already leaned conservative or considered themselves independent moved further right still—the algorithm amplified existing leanings rather than creating them from scratch. This distinction matters. It suggests that algorithmic feeds do not simply radicalize everyone equally; they intensify the political views people already hold, creating a kind of gravitational pull toward the ideological edges.

Ramón Salaverría, a journalism professor at the University of Navarra and coordinator of Iberifier, an Iberian digital media observatory, praised the study's rigor. The research was conducted independently, without collaboration from X itself, which strengthened its credibility. Celia Díaz, a researcher at the TRANSOC Institute at the Complutense University of Madrid, called it "a rigorous warning about the plasticity of political attitudes in the face of digital platform architecture." She noted that unlike earlier studies conducted with Meta, this research measured both the activation of the algorithm—switching from chronological to algorithmic mode—and its deactivation, capturing the persistence of algorithmic effects in ways previous work had missed.

The timing of this research is significant. It arrives as governments and regulators worldwide grapple with how to govern social media platforms, and as these platforms have become the primary news source for millions of people. The study demonstrates that algorithmic mediation is far from neutral. It does not simply organize information; it shapes the political landscape users inhabit. Belén Laspra, a philosophy professor at the University of Oviedo, called it "one of the most solid experimental pieces of evidence to date on the relationship between recommendation algorithms and political attitudes on social media." In an era defined by intense debates over misinformation, polarization, and algorithmic governance, this kind of rigorous evidence has become essential to understanding how the digital platforms we use daily are reshaping our politics.