Over the past decades, the diffusion of fake news and other deceptive content on social media platforms has become a heated topic of debate. Some past studies have explored the broad impact of online misinformation, while others have tried to determine whether deceptive content influences voters during political elections.
Researchers at Stanford University, Meta, University of Pennsylvania and other institutes recently set out to shed light on the prevalence of deceptive networks on social media in 2020, at the time of US presidential elections. Their paper, published in Nature Human Behaviour, offers interesting insight about 49 deceptive online networks that reached millions of Facebook and Instagram users during the 2020 elections.
“This paper grew out of the US 2020 Facebook and Instagram Election Study, a collaboration between academic researchers and Meta to study the political effects of social media around the 2020 elections,” Jennifer Pan, co-author of the paper, told Phys.org.
“Much of the existing research on influence operations has focused narrowly on foreign operations—especially Russian ones around 2016—and has had to rely on proxies for exposure, like whether someone followed or mentioned a network account, rather than whether they actually saw the content.”
Analyzing data collected by Meta
Building on earlier studies focusing on deceptive online content, Pan and her colleagues set out to identify the organized groups of online accounts that disseminated misleading political information during the 2020 US elections. They focused on both disincentivized networks of users who engaged in inaccurate political discourse and financially motivated networks disseminating content that is largely dismissed as spam or clickbait.
“We also wanted to measure actual exposure—how many real users saw this content—which is something only platforms can see,” said Pan. “Our central question was straightforward: how do deceptive online networks actually reach audiences, and who ends up being reached?”
As part of their study, the researchers closely analyzed 49 deceptive networks that targeted adult Facebook and Instagram users in the US during the 2020 election. These networks were identified by researchers at Meta based on data collected by the company.
13 out of the 49 identified networks were found to be what Meta refers to as coordinated inauthentic behavior networks. These are essentially networks of accounts that post deceptive content aimed at influencing the political views of social media users.
The remaining 36 networks were found to be financially motivated. This means that the posts they disseminated were designed to invite users to click on them and produce revenue.
“For each network, we looked at three things: their characteristics (where they originated, how many accounts they ran, what they posted about), their activity on the platforms, and their reach—meaning the number of unique US adult users who actually saw their content, either directly from a network account or indirectly when someone unaffiliated with the network reshared network content,” explained Pan.
“We also ran a complementary analysis on about 73,000 consenting survey participants, linking their survey responses to their on-platform exposure, so we could explore associations with outcomes like the ability to tell true from false information.”
A deeper understanding of deceptive networks in 2020
The researchers found that the 49 deceptive online networks identified by Meta reached at least 37 million Facebook users and 3 million Instagram users over the eight-month window they examined. This means that misleading content reached approximately 15% of active Facebook users in the US.
Perhaps even more strikingly, Pan and her colleagues found that the reach of the networks was highly concentrated. In fact, only 3 of the 49 networks accounted for over 70% of all the users reached across both Facebook and Instagram, one of which was an account called “Rally Forge’ created in the US.
“We also showed that the mechanism of reach matters,” said Pan. “On Facebook, networks reached most of their audience not directly, but because ordinary users—people unaffiliated with the networks—reshared their content. The network with the highest reach, for example, reached about 1.3 million users directly, but 13 million indirectly through reshares by ordinary users.”
Interestingly, the researchers observed that financially motivated networks, which some previous studies dismissed or considered less impactful in the context of elections, produced a substantial amount of political content. Moreover, the content they disseminated often reached far more users than the posts shared across politically motivated networks.
“One thing worth emphasizing is that the exposure data came from platform-level measurements—these are actual views, not inferred exposure—which is fairly unusual for this kind of research,” said Pan.
The team’s analyses also gathered information about the demographics of users who were more exposed to deceptive networks in the months leading up to the 2020 elections. Specifically, they revealed that users reached by the networks were typically middle-aged or older, more active on Facebook, more likely to engage with untrustworthy content and politically conservative.
Implications and next research steps
This recent study offers some valuable insight into how political misinformation spreads online and what types of users tend to be more exposed to deceptive content. Overall, it suggests that fake news or other misleading posts spread not only via its creators, but also because they are widely shared by regular users.
The results of the team’s analyses could soon inform the development of new countermeasures designed to detect and reduce misinformation on social media platforms. For instance, they suggest that interventions that only target deceptive networks might be insufficient, as regular users are also contributing to the dissemination of misleading content.
“If a small number of non-network users are doing most of the amplification, then understanding why they reshare—and designing interventions around that behavior—becomes just as important as counteracting the networks themselves,” said Pan. “This also means that platforms need to take financially motivated operations seriously as part of the political information environment.”
As a next step, Pan believes it would be important to better understand what type of users tend to re-share deceptive online content and why they choose to do so.
“These users are a small share of those exposed, but they drive most of the downstream reach, and we still know relatively little about their motivations,” added Pan.
“Another future direction would be to focus on platform features—how specific design choices, like reshare mechanics, enable or constrain the spread of deceptive content. More broadly, all of this depends on continued platform transparency and data access for researchers, which has decreased greatly in the past six years.”
Written for you by our author Ingrid Fadelli, edited by Gaby Clark, and fact-checked and reviewed by Robert Egan—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive.
If this reporting matters to you, please consider a donation (especially monthly). You’ll get an ad-free account as a thank-you.
Publication details
Ruth E. Appel et al, How deceptive online networks reached millions in the US 2020 elections, Nature Human Behaviour (2026). DOI: 10.1038/s41562-026-02435-2.
The content is provided for information purposes only.