X Fact-Checks Don't Stop Repeat Misinformation Posters
Fact-check labels may be backfiring on the worst online offenders.
Community Notes is X's crowdsourced fact-checking system, where users add context to misleading posts. A new study tracked nearly 20,000 accounts for four weeks before and after their posts got labeled, using 57,935 corrections and over 11 million posts. The goal was simple: do these public labels actually change how people post afterward?
The answer is mixed. On average, corrected accounts posted 2.9% more original content after being flagged. But that average hides two very different groups. Accounts corrected only once actually posted 2.4% less, and their new posts were less toxic and less misleading. In other words, a single public correction seemed to nudge them toward better behavior.
The problem is the other group. Repeatedly corrected accounts—just 28.6% of all corrected accounts, yet responsible for 73.4% of fact-checked posts—did the opposite. After their first correction, they posted 4.4% more. Later corrections had no effect at all. They didn't improve their content; instead, they shared more highly misleading posts, cited lower-quality sources, and posted more political content.
The researchers conclude that correcting content and changing the people who create it are two separate challenges. Community Notes can constrain an individual false post, but it doesn't reform the worst repeat offenders. For platforms designing against misinformation, this suggests labels alone aren't enough—you need features that directly target account behavior, not just individual posts.
- One-time offenders post 2.4% less and clean up their content after being fact-checked
- Repeat offenders make up 28.6% of corrected accounts but produce 73.4% of flagged posts
- Repeat offenders post 4.4% more after first correction, with more misinformation and political content
Why It Matters
Social media fact-checking may need to target repeat misinformation accounts, not just individual false posts.