Imagine trying to drain a swamp by scooping out the alligators one at a time. You might remove a few, feel productive, even celebrate the empty spots—but the water, the weeds, and the hidden channels underneath remain, ready to fill the space again. That, in a nutshell, is what new research from George Washington University suggests about the fight against online health misinformation. For years, the standard response to dangerous falsehoods has been to delete anti-vaccine posts and pages, ban repeat offenders, and spend millions on fact-checking, moderation, and public education campaigns. According to a sweeping new study of Facebook communities, however, this approach has made surprisingly little difference to the structure of the misinformation network itself. The researchers mapped 1,356 interconnected Facebook communities between 2019 and 2025—an ecosystem representing an estimated 70 million to 100 million combined page memberships. They watched what happened as Facebook removed pages and connections, as moderation budgets grew, and as public awareness of misinformation increased. The network, they found, is still largely intact. Even more troubling, the distrust that once centered on vaccines has now drifted into entirely new territory, including elections in the United States and beyond. In other words, the same social architecture that helped anti-vaccine sentiment flourish has become a general-purpose conduit for political distrust. The findings, published in the Journal of Complex Networks, challenge the assumption that misinformation is a problem of scattered bad actors who can be removed one by one, rather than a deeply rooted social ecosystem that adapts, absorbs shocks, and grows.
The numbers are sobering. During the study period, the overall network lost roughly 28 percent of its pages and 47 percent of its connections. Anti-vaccine communities took an even bigger hit, losing 33.5 percent of their pages and 52.3 percent of their connections. If deletion were the answer, this should have been a decisive blow. Instead, the remaining communities simply tightened their ties to a broader web of distrust. Neil Johnson, a professor of physics at George Washington University and one of the study’s authors, puts it plainly: “Removing individual pages and connections does not dismantle the networks that allow misinformation to spread.” The reason, he explains, is that these communities are not isolated bubbles. They are connected to larger, sprawling networks that cross topics and geographic boundaries. A small anti-vaccine page in one country might be linked to a parenting group, an alternative-health forum, a climate-skeptic page, and an election-integrity group in another country. When Facebook removes the page, the surrounding network remains, and its members quickly form new connections elsewhere. Think of a highway system: closing one exit doesn’t stop traffic; drivers simply find another route. Similarly, when a prominent anti-vaccine page disappears, its followers don’t vanish with it. They migrate to smaller pages, private groups, or pages that don’t look political at all. The network’s resilience is not just a matter of stubborn users; it is built into the way communities form and connect.
So why are these networks so difficult to disrupt? The researchers point to two key features. First, communities increasingly talk about multiple topics at once. A page that starts with vaccine concerns may also post about parenting, alternative health, climate change, and elections. This cross-topic blending means that misinformation no longer stays in a single lane. Someone who arrives curious about childhood vaccines might be swept into a much broader worldview of distrust—one that casts doubt on public health, scientific expertise, and political institutions alike. The researchers call this the “distrust ecosystem,” and they argue that understanding it requires an updated map, not just a list of banned pages. Second, these communities maintain strong connections between local groups and international networks. A local parenting group in Ohio can be linked to a global anti-vaccine movement through a shared post, a common administrator, or a similar hashtag. This local-global bridge gives the network enormous flexibility. When one national page is removed, its message can still flow through many smaller, harder-to-detect channels. As Johnson and his colleagues write, the updated map is essential because “it tells us not just what people see online, but how the communities around them are connected.” In other words, the most important unit of analysis may not be the individual post or page, but the dense web of relationships that gives misinformation its staying power. You might look at a single suspicious post and wonder why anyone would share it, but to understand why it spreads, you have to look at the neighborhood it lives in—and all the invisible roads leading in and out.
Perhaps the most important takeaway is practical: continuing to simply remove individual posts or pages is not enough to stop misinformation from spreading and distrust from growing at the population level. To test what might work better, the researchers used their updated distrust map to build computer simulations. They modeled online discussion groups populated by people with different perspectives—pro-vaccination, anti-vaccination, and neutral—and then watched what happened when those groups interacted. The results were striking. When the simulated communities engaged in genuine deliberation, the share of anti-vaccine communities decreased over time from 34.2 percent to 26.5 percent, while the share of neutral communities increased from 47.9 percent to 65.4 percent. In other words, simply bringing people together to talk—rather than shouting past each other or trying to silence one side—seemed to shift the balance away from distrust. These findings are backed, the researchers note, by their prior smaller-scale testing with human communities, which showed that encouraging constructive dialogue between sets of online communities could significantly outperform content moderation alone. This is not an argument for leaving harmful misinformation alone. It is an argument for changing the battlefield. Instead of playing an endless game of whack-a-mole with pages and posts, platforms might invest in tools that help communities interact across lines of disagreement, create spaces for nuanced conversation, and foster the kind of long-term trust that makes simplistic conspiracy theories less appealing. The underlying idea is hopeful: people are not permanently trapped in extreme views. When given respectful opportunities to share concerns and hear different perspectives, many will soften their positions or at least become open to more balanced information.
At the same time, it is important to understand what this study did and didn’t do. The researchers examined publicly visible Facebook Pages, not personal accounts or private groups. This means the true scale of these distrust networks may be even larger than what the data show, since private groups are often where the most extreme content thrives. It also means the findings are not a universal law of human psychology, but rather a detailed snapshot of one major platform during a turbulent period of history. Even with that caveat, the implications are hard to ignore. By tracing how communities connected and shifted over six years, the study sheds light on the current spectrum of distrust at a population scale—not just in the United States, but around the world. It also reveals how new forms of online misinformation and polarization emerge. Vaccine hesitancy, the researchers observe, was never really just about vaccines. It was a doorway into a broader way of seeing the world, one that could later accommodate election denial, climate skepticism, and distrust of institutions in general. Understanding this pathway is crucial. It means that when public health officials respond to a vaccine scare by offering more facts and figures, they may be addressing the wrong layer of the problem. The challenge is not just correcting a single false claim; it is engaging with a social identity, a community, and a shared sense of who to trust. A person who has come to believe that experts are lying about vaccines is unlikely to be convinced by another expert. But they might be willing to listen to a neighbor, a faith leader, or a fellow parent who shares their values and acknowledges their fears.
Ultimately, the study offers both a warning and a possibility. The warning is that the misinformation ecosystem is more resilient than we thought, and that content moderation alone—no matter how aggressive—will not be enough to rebuild public trust. The possibility is that the same network structures that spread distrust can also be used to spread something else: dialogue, understanding, and connection. The simulations suggest that when online communities are deliberately brought into conversation, the share of neutral voices grows and the share of anti-vaccine voices shrinks. This is not a quick fix or a magical formula; it is a slower, harder, and more human process. It requires platforms to think of themselves not just as content hosts, but as facilitators of relationships between communities. It requires policymakers and public health leaders to see the social landscape as a living network, not a collection of isolated bad apples. It requires all of us to recognize that behind every extreme online community is a person searching for certainty, belonging, and a sense that someone takes their concerns seriously. The challenge, as the George Washington University researchers put it, is to update our distrust map—and then to use it to build bridges rather than just tear down walls. If we can do that, the same architecture that once helped misinformation spread across borders and topics might one day be used to help wisdom, compassion, and trust do the same.