In case you had any doubt, Elon Musk’s X has an algorithm that favors conservative content posted by political activists over liberal content or posts by traditional news media accounts, according to a new study published in Nature.

X, then-Twitter, had a chronological timeline until 2016, when the company decided to give users the option to switch to a feed curated for them by an algorithm. After Musk acquired the company in 2022, the algorithm-defined “For You” page became the default page, with a tab for people that users follow. And recently, the algorithm has become “purely” AI-led, according to Musk, the AI in question being the controversial Grok, aka MechaHitler.

Over the past few years, many have accused X’s algorithm of favoring right-wing content. A Sky News investigation from last year claimed that the algorithm boosted right-wing and extreme content, and posts made by politicians that Musk likes. The effect was more pronounced in accounts already somewhat subscribing to those beliefs, but was visible across the political spectrum.

Earlier this month, French police raided X’s offices in the country as part of an almost year-long investigation into a potential algorithm manipulation on the platform meant to “serve a political agenda.” Musk has called the probe “politically motivated.”

Wednesday’s new academic study, conducted by a team of researchers from Italy’s Bocconi University, Switzerland’s University of St. Gallen, and France’s Paris School of Economics, now gives a structured look into the phenomenon and how it might be leading to sustained change in the ways users consume media.

By monitoring the “For You” feed, the researchers found that not only was the algorithmic feed more engaging, it also favored conservative content over liberal content and posts by political activists over news media.

Posts that were conservative in nature were roughly 20% more likely to appear in an algorithmic feed, while those that were liberal were only 3.1% more likely. The contrast was even more pronounced when it came to the algorithm’s treatment of news. Traditional news appeared roughly 58% fewer times in users’ algorithmic feeds, while posts by political activists appeared 27.4% more often and those by entertainment accounts 21.5% more often.

The content pushed by the algorithm ended up impacting the way users think about some political issues, the researchers found.

For the study, researchers questioned a pool of U.S.-based X users on their political beliefs and how they use the platform. Then they randomly assigned each user to either a chronological or an algorithmic feed for seven weeks.

At the end of the experiment, those who were switched from the chronological feed to the algorithmic feed for seven weeks reported a conservative shift in political opinions. But there was no change in perspective for those who were previously using the algorithmic feed and were switched to the chronological one.

Specifically, the impacted users reported caring more about conservative policy priorities like immigration, and supported a more negative view of the multiple criminal investigations opened against Trump in 2023 and a more pro-Kremlin view of the Russia-Ukraine war. This happened despite the users not reporting a change in party affiliation or level of affective polarization, a key term used by political psychologists to define a deep-seated animosity against groups in opposition to the one you subscribe to.

The rightward shift in opinion that accompanied the switch to the algorithm-based feed is also likely to last even after the experiment is over, researchers concluded. That’s because the users who switched to the algorithm ended up following more conservative political activist accounts, while there was no significant difference in the likelihood to follow liberal accounts or news outlets.

“Our findings demonstrate that the algorithm significantly influences whom users choose to follow, indicating that algorithms shape content exposure more than believed previously,” the study claims, adding that it can have “a lasting impact on users’ feeds and their political attitudes.”