When Elon Musk took over Twitter and renamed it X, one of his most ambitious ideas was Community Notes. It was pitched as a kind of digital democracy in action: instead of relying on a central authority to decide what was true or false, the platform would let everyday users write short notes underneath posts to add context, correct mistakes, or flag misinformation. The system sounds appealing on its face. It suggests that the truth can emerge from the collective wisdom of millions of people, almost like a conversation correcting itself in real time. And it has been praised by Musk himself, who has called it “awesome.” But a new study from NewsGuard, a US watchdog that rates the credibility of news and information sites, has cast serious doubt on whether this crowdsourced approach actually works when it matters most. According to the study, released on October 21, Community Notes failed to flag more than 93 percent of false claims related to the US midterm elections, including false claims promoted by Musk himself. In other words, the very tool that was supposed to help X combat misinformation is missing the vast majority of it during one of the most consequential election cycles in the country. The findings are a sobering reminder that crowdsourcing the truth is far more complicated than it sounds.
The NewsGuard study focused on posts that advanced false claims about the midterm elections, including bogus allegations of election fraud and questionable voting rules. Researchers identified dozens of such posts on X, and then checked which ones had been tagged with Community Notes. The results were striking: only 13 of those posts had any note attached at all. That means 93.1 percent of the false claims reviewed by NewsGuard were left completely uncorrected, free to circulate, accumulate likes, and shape public perception. Perhaps even more concerning, these uncorrected posts had already drawn millions of interactions ahead of the November 3 midterm elections. One example of the false content involved a conservative commentator who claimed that New York and California had made it illegal for voters to show their voter ID, and also illegal for election officials to ask for one. That is simply not true. No such rule exists in either state, both of which are Democratic strongholds. But without a Community Note to correct it, the post could easily appear credible to someone scrolling quickly through their feed. This is exactly the kind of falsehood that Community Notes was designed to catch: a clear, factual error with real-world consequences. The fact that it slipped through, along with so many others, raises serious questions about whether the system can be relied on as a meaningful safeguard against election misinformation.
Perhaps the most uncomfortable part of the study is what it says about Musk himself. Musk has often promoted Community Notes as a model of transparency, saying that all users, including himself, are subject to correction by the crowdsourced system. He has presented it as a form of accountability that applies equally to everyone. But NewsGuard found that Musk, who has 241.5 million followers on X, advanced at least four false claims related to the midterm elections, and none of them carried Community Notes. That means the most influential account on the platform has been able to spread false information without any visible correction, despite Musk’s own claims that the system holds everyone accountable. This is not just a theoretical concern. When someone with that kind of reach posts something misleading, it can instantly reach millions of people. A Community Note added hours later, if it is added at all, may do little to undo the damage. X did not respond to AFP’s request for comment about the findings. But the silence itself is telling. If a major study reveals that the platform’s flagship misinformation tool is failing, users have a right to expect a response. Instead, the question hangs in the air: if Community Notes cannot catch false claims made by the owner of the platform, what is it actually catching?
Researchers who have studied Community Notes say that the system can work in certain contexts, but it is far from a universal solution. Studies have shown that it can be effective in dispelling some types of falsehoods, such as vaccine misinformation, where there is broad scientific consensus and relatively little partisan division. When almost everyone agrees on the facts, crowdsourced correction tends to work well. But the story is very different when it comes to highly polarized topics like elections. In those cases, the people who are most motivated to write notes may be driven by partisan beliefs rather than a genuine desire for accuracy. Some researchers have cautioned that Community Notes can be manipulated by users who coordinate to target political opponents, downvoting notes they disagree with and upvoting notes that align with their side. The system relies on a belief that the crowd will be fair, but crowds can be biased. They can be organized. They can be angry. And when the topic is something as emotionally charged as an election, the crowd may not be looking for truth at all. This is why professional fact-checking organizations have traditionally played such an important role. They are trained to verify claims, follow evidence, and correct the record regardless of political pressure. Crowdsourcing, by contrast, can feel more democratic, but it is also more chaotic.
Despite these concerns, tech platforms are increasingly embracing the community-driven model as an alternative to professional fact-checking. The most significant example of this shift came in 2025, when Meta announced it was ending its third-party fact-checking programme in the United States. CEO Mark Zuckerberg said the programme had led to “too much censorship,” a claim that professional fact-checkers have vehemently rejected. In the place of independent fact-checking, Meta has begun testing its own version of Community Notes, first in the US and more recently in 16 Spanish-speaking countries in Latin America. Meta has not provided a timeline for when the change will become official, but the direction is clear: the company is moving away from professional oversight and toward user-generated corrections. This has raised alarm among advocates for media literacy and election integrity. Even Meta’s own Oversight Board, a quasi-independent body that the company funds to review its moderation decisions, has said that Community Notes is not an adequate replacement for fact-checking. The board’s warning is significant because it comes from a body that is deeply familiar with the complexities of content moderation. It recognizes that while community notes can complement fact-checking, they cannot replace the rigorous, ethical standards that professional fact-checkers bring to the table.
As the US midterm elections approach, the failure of Community Notes to flag more than 93 percent of false claims is not just a technical problem. It is a human problem. It is about voters trying to make sense of a confusing and often misleading information environment. It is about the difference between a platform that takes its responsibility seriously and one that simply hopes the crowd will fix things. The idea behind Community Notes is appealing because it trusts ordinary people to participate in the pursuit of truth. But the evidence increasingly suggests that trust alone is not enough. False information spreads quickly, especially when it is repeated by influential figures, and the mechanisms designed to slow it down are often too slow, too fragmented, or too easily manipulated. The watchdog NewsGuard’s findings are a reminder that crowdsourcing and professional fact-checking should not be seen as competing approaches, but as complementary ones. A note from a fellow user can be helpful, but it is not a substitute for an independent, accountable fact-checking process. AFP currently works in 26 languages with Meta’s global fact-checking programme, including in Latin America, and that kind of collaboration is more important than ever. In the end, the question is not whether Community Notes is “awesome” in theory. The question is whether it can be trusted in practice, when the truth really matters. And based on the evidence, there is still a long way to go.