The pervasive worry that social media is a hotbed of toxic misinformation—and that cleaning up our feeds is the magic bullet for a healthier society—might be based on a misunderstanding of how we actually consume information. A major collaborative study, involving researchers from prestigious institutions like Stanford, Princeton, and NYU, recently challenged these assumptions. Over a three-month period during the intense 2020 US election, researchers cut exposure to content from untrustworthy sources by roughly 70% for over 15,000 users. Surprisingly, the data showed that this massive reduction had virtually no impact on people’s actual beliefs, their trust in mainstream media, or the depth of their political polarization. Even for those who were previously “power users” of misinformation, scrubbing their feeds did not move the needle on their worldview.
This lack of movement suggests that our beliefs are far more deeply rooted than a single post or a doctored image might lead us to believe. The authors of the study, published in Science Advances, noted that while it is technically feasible to filter out “untrustworthy” sources, such interventions are unlikely to produce an immediate change in public opinion. It seems that by the time a person is engaging with questionable content, their ideological leanings are already well-established. This pushes us to reconsider the role of algorithms; if the content itself isn’t the primary driver of radicalization, then our focus on simply removing the content might be a case of treating the symptom rather than the disease.
Perhaps even more surprising was the study’s findings on the actual volume of misinformation people see. While we often imagine our digital landscape as being overwhelmed by fake news, the numbers paint a different picture. For the average user, content from “untrustworthy” sources—defined by researchers as those with multiple fact-checking strikes—accounted for only 1.1% of their Facebook feed and a microscopic 0.1% on Instagram. The idea that the average person is being bombarded by misinformation 24/7 appears to be a perception gap; most users are simply not seeing as much of it as we assume, suggesting that the “misinformation crisis” is highly concentrated rather than widespread.
However, this exposure is not distributed equally. The study revealed a “heavy user” phenomenon, where a small fraction of the population accounts for the vast majority of misinformation consumption. About 23% of Facebook users and 11% of Instagram users were responsible for roughly 80% of all untrustworthy content exposure. For the top 2.5% of these users, misinformation made up a significant portion of their daily diet. What makes this even more telling is that this exposure wasn’t primarily the result of mysterious, radicalizing algorithms feeding them garbage. Instead, it was largely self-inflicted; these users were consuming this content because they had chosen to follow those sources directly.
The study serves as a poignant reminder of the limitations of corporate intervention. As Meta moves away from its third-party fact-checking strike system toward a community-notes model, the specific tools used in this experiment are being retired. The researchers warn that we are stepping into a period of uncertainty. Without the enforcement mechanisms that once allowed researchers to track the flow of information, it becomes nearly impossible to evaluate how these changes affect our digital environment. The authors emphasize that we are effectively flying blind; without rigorous, transparent scientific data, policymakers and tech companies are left to guess at the societal impact of their changing moderation policies.
Ultimately, this research invites us to take a step back from the panic surrounding social media. It suggests that while the digital world is certainly messy and filled with problematic corners, simply turning the dial down on misinformation is not a substitute for meaningful conversation or critical media literacy. The fact that removing the content didn’t change the minds of the people consuming it proves that our digital lives are complex, and the solutions to our deep-seated political divisions are likely much more difficult to implement than just hitting a “block” or “remove” button. As we move forward into a new era of online moderation, the biggest takeaway may be that our shared reality is shaped far more by our personal choices than by the content the platforms try to hide from us.