New research examining online behavior during the 2024 US presidential election suggests that social media algorithms can measurably influence how people view supporters of the opposing political party.

A team of US academics has developed an AI-powered browser extension that lets independent researchers analyze how social media feeds shape political perceptions, without needing cooperation from platforms like X. The findings indicate that reducing the visibility of hostile or extreme political content could help ease partisan divides.

The tool scans a user’s X feed for content containing anti-democratic themes or highly negative partisan messaging, including posts that imply violence or call for prosecuting political opponents. Using AI, the extension automatically reorders the feed within seconds, pushing the most polarizing posts farther down.

Researchers from Stanford University, the University of Washington, and Northeastern University tested the extension on more than 1,200 volunteers over a 10-day period leading up to the 2024 election. Some participants saw feeds where polarizing content was elevated, while others saw versions where that content was pushed lower.

Small algorithm tweaks, noticeable attitude shifts

Participants were asked throughout the study to rate their feelings toward the opposing party on a 100-point scale. Those using the version that down-ranked hostile content became, on average, two points warmer toward the opposing side, a shift researchers say is equivalent to three years of typical change in public attitudes.

“These changes were comparable in size to 3 years of change in United States affective polarisation,” the authors wrote.

The shift appeared on both sides of the political spectrum, affecting liberals and conservatives alike.

'Clear' impact on polarization, researchers say

Tiziano Piccardi, an assistant professor of computer science at Johns Hopkins University, said the intervention had a “clear” effect.

“When the participants were exposed to less of this content, they felt warmer toward the people of the opposing party,” he said. “When they were exposed to more, they felt colder.”

The team argues that their browser extension demonstrates a way to study, and even reshape, online discourse “without platform collaboration.”

“These interventions may result in algorithms that not only reduce partisan animosity but also promote greater social trust and healthier democratic discourse across party lines,” the study concluded.

The researchers also measured emotional responses. Participants who saw fewer hostile posts reported feeling less anger and sadness while browsing, though these emotional improvements did not persist after the experiment ended.

The tool only works on X accounts accessed via desktop browsers, not through the mobile app, an important constraint given how many users rely on smartphones. The researchers also note that they did not examine long-term impacts, leaving open the question of whether reduced exposure to polarizing content can generate durable shifts in political attitudes.