A new academic study shows that Community Notes, the crowd-driven fact-checking system on X, sharply reduce the reach of misleading posts. Researchers tracked tens of thousands of posts over a four-month period and found that when notes were attached, users were less likely to share or endorse the content.
Clear Reductions in Engagement
The team analyzed more than 40,000 posts proposed for notes between March and June 2023. Around 17 percent of them received approved notes. Using synthetic control methods, the researchers compared engagement on posts with notes against similar posts without them.
The numbers showed a consistent drop. In the 48 hours after a note was added, reposts fell by about 46 percent, likes dropped by 44 percent, replies by 22 percent, and views by 14 percent. Looking at the posts across their entire lifespan, the effect was smaller but still significant, with reposts down 12 percent and likes down 13 percent on average.
How Information Cascades Change
The study also reconstructed repost chains to examine how misinformation travels. When a note was attached, repost cascades became shorter and less viral. Posts were less likely to spread through long chains of person-to-person sharing. This shift meant that people further from the original poster were less exposed to the misleading content.
The effect was most visible on users who came across a post through several layers of sharing. Followers close to the original author were less influenced, suggesting that social proximity plays a role in how users respond to corrections.
Timing Is Critical
One of the clearest findings was the importance of timing. Notes attached within the first 12 hours of a post being published were far more effective. Posts corrected quickly saw large drops in reposts and likes. When notes appeared after two days, the impact was minimal. In some cases, attaching late notes even drew renewed attention to old posts, though this did not translate into stronger endorsement.
Stronger Effects on Media Content
Posts containing images or video saw larger declines in sharing than text-only posts. Notes that flagged altered media or misleading satire also proved more effective than those correcting simple text errors.
The researchers found that note style mattered. Notes written in clear, simple language had greater impact than complex ones. Moderately long notes, usually two to three sentences, performed better than either very short or overly detailed notes.
Limits of the System
While Community Notes clearly reduce engagement, their effect is uneven. Views fell by only 5 percent overall, and replies by about 7 percent. This means users can still be exposed to misinformation even when it carries a correction. Exposure alone can influence later beliefs, which is why the researchers view each prevented view as valuable.
The study also highlighted limitations. Only posts that receive notes can be measured, leaving open the question of how much misinformation goes unchecked. The researchers also cautioned that the system’s impact may change over time as X adjusts how notes are ranked and displayed.
Wider Context
X introduced Community Notes in 2022 as its main fact-checking system, after reducing its content moderation staff. The system relies on ordinary users to write and rate notes, with an algorithm ensuring that only broadly supported notes are displayed. Other platforms such as Meta and YouTube have introduced similar features, aiming for scalable alternatives to professional fact-checking, which is slower and harder to expand.
Future Outlook
Since the study period, X has made technical changes that may improve the tool. Updates include faster processing of notes and automatic links between new posts and previously flagged material. These refinements are expected to increase the speed of corrections, which the research identified as the most important factor for effectiveness.
The findings suggest that Community Notes can act as a meaningful check on misinformation, especially when deployed quickly and on posts containing visual content. But the authors also note that no single measure is sufficient, and that a combination of tools remains necessary to limit the spread of false information online.
Notes: This post was edited/created using GenAI tools.
Read next: TikTok’s Shopping Test Raises Concern After Gaza Clips Flagged
