A Science Advances study run with Meta during the 2020 U.S. election found that exposure to untrustworthy sources was concentrated among a small minority of users, and that reducing it left attitudes unchanged.

For years, social media companies have faced pressure to curb the spread of misinformation, on the theory that false and misleading content can distort elections, deepen political divisions and erode trust in institutions. The assumption behind much of that pressure is simple: if people see less misinformation, they will believe less of it.

A new study tests that assumption directly, and the results are more complicated than many expected. Writing in Science Advances, researchers report that sharply reducing Facebook and Instagram users’ exposure to content from untrustworthy sources during the 2020 U.S. presidential election did not measurably change their political beliefs or attitudes.

“The intervention was successful in the sense that it reduced what it was supposed to reduce, in this case, exposure to information from untrustworthy sources,” said Olivier Bergeron-Boutin, a Ph.D. student in political science at the University of California, Berkeley, who led the study. “But the resulting changes in attitudes were much smaller than some would have expected.”

An experiment inside the feed

The study was part of a large research project in which independent academics worked with researchers at Meta, the parent company of Facebook and Instagram, during the 2020 election. Earlier papers from the same collaboration, published in Science and Nature in 2023, tested changes such as switching users to chronological feeds and reducing content from like-minded sources, and they too found little effect on political attitudes.

In this experiment, consenting Facebook and Instagram users were assigned either to their normal feeds or to feeds that showed less content from sources that had repeatedly spread misinformation. According to the release, the intervention relied on Meta’s fact-checking “strike” system, which tracked sources flagged repeatedly by third-party fact-checkers.

Over three months around the election, the intervention worked as designed. Participants in the treatment group saw about 70 percent less content from untrustworthy sources. Yet their political beliefs and attitudes showed no measurable change, even among people who had previously seen the most content of this kind.

A concentrated problem

The study also found that exposure was highly uneven. Most users saw little content from untrustworthy sources. About 23 percent of Facebook users and 11 percent of Instagram users accounted for roughly 80 percent of all exposure.

“Exposure to content from untrustworthy sources on social media is less common and more concentrated than people typically think,” said co-author Brendan Nyhan, the James O. Freedman Presidential Professor at Dartmouth College. “In addition, reducing exposure to content from those sources did not have the effects that many speculated it would have on people’s beliefs, attitudes and behaviors.”

The study’s other authors are Jaime Settle of William & Mary, Emily Thorson of Syracuse University and Magdalena Wojcieszak of the University of California, Davis.

Why beliefs don’t budge

Tim Tangherlini, a computational folklorist at UC Berkeley who was not involved in the study, offered one explanation. “Stories don’t spread evenly through a population,” he said. “They flow on and across social networks predisposed to receive them, retell them and refine them — basically like an echo chamber.”

From that perspective, cutting off one source of content may not be enough to change what a community believes. “A platform intervention changes what flows into the network,” Tangherlini said, “but it doesn’t change the underlying social network, or the beliefs shared in that network.” According to the release, he compared changing beliefs to steering a large tanker, which takes time.

Political attitudes are also known to be relatively stable. Partisan identity, in particular, tends to be formed early and to shift slowly, which may make it resistant to a few months of altered news feeds.

Focus on the heavy users

Nyhan argues that the findings should change where attention is directed. “I hope social media companies and policymakers will recognize that exposure to content from untrustworthy sources is heavily concentrated and the harms from it are likely to be similarly concentrated,” he said.

“We should worry less about the average person (who is typically uninterested in politics) and more about potential effects on people who consume large volumes of dubious content online and what it might do to them or inspire them to do.”

Bergeron-Boutin cautions against reading the results as a clean bill of health for misinformation. “We shouldn’t interpret these results to mean that misinformation is no big deal,” he said. “But there’s no quick fix. Anyone who proposes a simple solution to misinformation or political polarization should be viewed skeptically.”

“Tinkering at the margins can help, but ultimately, institutional features of American democracy provide fertile ground for these issues to rise,” he added.

A study that may not be repeated

The researchers stress that their data describe 2020, not today. In 2025, Meta ended its third-party fact-checking program in the United States, shifting toward a crowd-sourced Community Notes system and removing the strike system that made this kind of source-level intervention possible. The findings, they say, cannot describe exposure to misinformation on the platforms now.

The authors also worry about future research. “This study presents a unique opportunity because social media platform data is rarely accessible to outside researchers,” Bergeron-Boutin said. “A study like this is becoming less and less likely to happen again, because since 2020, these platforms have become less transparent.”

The release does not report the number of participants. Nyhan sees the broader task as finding better tools. “We need to learn how to better measure the potential harms from dubious content online and address them without compromising the values and freedoms that we hold dear,” he said.

---

Notes

1. Bergeron-Boutin, O., Settle, J., Thorson, E., Wojcieszak, M., and Nyhan, B. “Untrustworthy sources on Facebook and Instagram in 2020: Concentrated exposure but no attitudinal effects.” Science Advances, 29 July 2026. DOI: 10.1126/sciadv.adz6502.
2. University of California, Berkeley. “Does misinformation on Facebook and Instagram change what you believe?” EurekAlert!, 29 July 2026. https://www.eurekalert.org/news-releases/1138146.

Photo by Pixabay on Pexels.com