Community Notes – X’s Crowdsourced Fact-Checking Feature
Community Notes is X’s flagship crowdsourced fact-checking feature, designed to provide timely, user-generated context on misleading posts. In practice, the system faces global challenges, particularly in high-risk languages such as Hindi and Urdu, where accurate notes often stall while misinformation remains visible. A study by the Center for the Study of Organized Hate (CSOH) reveals that posts in South Asian languages, including Hindi, Urdu, Bengali, Tamil, and Nepali, account for only 0.094% of the archive, even though these languages represent approximately a quarter of the world’s population and five percent of X’s monthly users. Only 37 of those notes have ever appeared on the public timeline, highlighting the system’s vulnerability of misinformation to systemic biases.
Key mechanisms:
The System relies on two main mechanisms to aggregate feedback: a “helpfulness” up.uploadment feature and a bridging test, which requires agreement from contributors who typically disagree with one another. The idea is elegant in theory: bring together people with different viewpoints to agree on what’s accurate, avoiding echo chambers. However, the system struggles in practice when there isn’t enough contributor activity in a given language to meet its consensus thresholds.
In the first chart, note-writing activity remained nearly flat until the Indian general election window from April to June 2024, when weekly volume briefly surged twentyfold. This pattern tells two stories: first, crises do mobilize contributors; second, the system isn’t designed to scale with that urgency. Drafts pile up precisely when voters most need real-time context.
Despite these challenges, its operations remain a concern for South Asian audiences, particularly those vulnerable to misinformation. The report tracks Community Notes authored in South Asian languages from April 2024 to April 2025 and reveals a stark trend. One chart shows note-writing activity remaining nearly flat until the Indian general election window, when weekly volume briefly surged twentyfold. This pattern tells two stories: first, crises do mobilize contributors; second, the system isn’t designed to scale with that urgency.
According to the study, posts in South Asian languages like Hindi occasionally rank higher, but misinformation remains visible. A key issue is reviewer scarcity: there are not enough raters who speak these languages and hold varied viewpoints, so the system cannot confirm the required cross-group agreement. Accurate notes often end up in draft form, while misinformation remains visible.
The hurdle is not just easier in Chinese or other widely spoken languages—it’s harder because it affects all South Asian languages. The report reveals that posts in South Asian languages have a higher rate of initial标记 than English, with South Asian landscapes having even higher variance in their initial rates. While Hindi and Urdu only have low marker rates, their posts fail to make it to the public timeline, despite contributing 35% of the overall archive. This imbalance reinforces existing linguistic fault lines and leaves South Asian audiences with less protection against misleading content.
Despite these challenges, some posts on South Asian languages rise to the occasion. For example, predictions about Indian Congress voters being “Italian-mafia supporters” in Hindi or slurswd terms against Pakistani police in Urdu were on the public timeline in July 2024. These drafts appear in draft form, with insufficient voting at submission, highlighting that discussions in South Asian contexts often misuse the feature.
Those submissions survive as drafts but receive littleinterpreted intent when made unintentionally. This is especially problematic in a region where the type of misinformation most dangerous involves political rhetoric. The data also reveal that these drafts not only bypass the feature’s moderation mechanisms but also replicate old problems: low success rates for low-frequency languages such as Bengali, which sends a信号 that.
When X initially rolled out the AI chatbots instead of notes, the system already struggled, according to its announcements. AI chatbots draft content in the middle of a series of crises, such as India’s elections earlier this year. X found itself in a maze of inequities, leaving毫升 for … but fails to announce context in Hindi, Urdu, Bengali, etc., as a feature designed to protect users.
To avoid this, X has to focus on a few things first—innocent steps. X must prioritize recruitment well before crises. For example, X opened note authorship in India only weeks before a national election. Platforms must cultivate contributor networks in high-risk languages year-round and conduct targeted recruitment drives six to eight weeks in advance of significant events that are foreseeable.
The system also must calibrate its publications thresholds to reflect linguistic realities. If a 1,000 reviewer group in Hindi provided sufficient vote totals to be seen as normal, its average would need to be at the consensus level for English. Without this adjustment, lower-ranked languages remain lost to higher-ranked ones.
Furthermore, platforms should implement a basic PAYLOAD filter at the point of submission. A lightweight language screen can block slurs and other abuses before drafts enter the ratings pool. While this improves readability and reduces participation in harmful discussions, it limits the feature’s ability to protect users from hate speech“I’m writing that because I can’t believe how the world is divided between India and Nepal. The central government is trying to implement policies that only harm migrants, and they’re not giving any proper support. But the usual percentage for protests against the COVID-19 pandemic is . . .”
Above all, platforms must meet the effort of introducing survey data on linguistic diversity. The $`next billion users’ communities are not phrased asavalanche of potential. If the industry treats multilingual trust and safety as an afterthought, we will have a two-tier internet. For an industry that prides itself on scale, this is both a market failure and a moral one. To build truly global communities, tech companies must prioritize linguistic equity in product design, as people will judge information quality based on how well all languages are supported.
Trust and safety for . goddess通知书dictate by the web’s next billion users. They will measure success in terms of coverage parity, not aggregate note count. If Tamil has approximately 80 million native speakers, the platform should be able to determine in real-time whether Tamil tweets receive proportional fact-check protection. Anything less is an illusion of integrity.
Conclusion:
X’s notes system remains a puzzle, with many in South India and Nepal sipping a glass of honey on the priority to fix reviewer pools. But less obvious is the swerves caused by the Leslikely of malicious, seeking to poison misinformation. Meta must make these steps first: prioritize recruitment before crises and adjust algorithm的专业ization to reflect linguistic realities. lt just cancels where it shouldn’t. Until then, X’s Modular toandering system is exactly that—a canister适合 its hollow life, designed to be both broke and broken.
