Sources and analysis: Twitter's Community Notes does not address divisive lies; over 30K, or ~96%, of notes contributed by the community are invisible to users
Bloomberg : Tweets: @daveyalba , @silviakillings , @sarahfrier , @kegill , and @daveyalba Tweets: Davey Alba / @daveyalba : “Abortion is never medically necessary to save the life of a mother,” reads a tweet from @FRCdc. This is factually untrue, but because more “left-leaning” accounts voted on the contextual note, the note is suppressed & Twitter says it “needs more ratings.” https://www.bloomberg.com/... https://twitter.com/... Silvia Killingsworth / @silviakillings : I'm shocked, shocked https://www.bloomberg.com/... Sarah Frier / @sarahfrier : We ran a data analysis of Twitter's Community Notes fact-checking program that Musk has touted. Turns out 96% of notes never make it to public view, since people on either side of the political spectrum can't agree: https://www.bloomberg.com/... by @fxy2014 @KurtWagner8 @rachaeldottle @kegill : This does not surprise me. It is frustrating to write a factual response with sources for obvious mis/disinfo and have it SIT. Too many writers haven't a clue how to provide fact-based content. Drives me more than a little nutz. https://twitter.com/... Davey Alba / @daveyalba : Let my brilliant colleagues @fxy2014 @rachaeldottle @KurtWagner8 walk you through how Community Notes *really* works https://www.bloomberg.com/... https://twitter.com/...
Context & Ripple Effects
Bloomberg's data analysis lands on a program that has been praised before: the Financial Times had called Community Notes a surprise hit for its transparent, crowd-sourced approach to fact-checking. The finding here is that the pipeline works but the output doesn't ship — over 30,000 notes, roughly 96% of what contributors wrote, never become visible, including corrections on divisive claims like a false statement about abortion and maternal health.
The visibility gap has proven durable. A volunteer contributor later described the tools as far from sufficient, and by late 2024 the Center for Countering Digital Hate found that of 283 X posts with false election claims that note-writers had responded to, only 20 carried publicly shown notes. Elon Musk has since pledged to fix the system, arguing it is being gamed, just as Meta begins rolling the model out.
First-order effects
- Contributors' labor is largely wasted: with ~96% of notes invisible, users seeing divisive false claims get no correction at all, so the note-writing community's effort produces almost no on-platform effect.
- Twitter carries the reputational cost directly — the system marketed as its answer to misinformation demonstrably fails on exactly the contested political claims it was designed for.
Second-order effects
- Meta, which is extending Community Notes across its services per Bloomberg's later reporting, inherits the same consensus-threshold bottleneck, meaning the flaw scales with the model rather than staying an X-specific bug.
- Musk's promise to fix notes he calls 'gamed by governments & legacy media' puts the platform's owner in open conflict with parts of the contributor base, risking contributor attrition on top of the visibility problem.
Third-order effects
- If the pattern holds, crowd-sourced fact-checking risks functioning more as a legitimacy shield for platforms than as an actual correction mechanism — a structural critique regulators and researchers are already documenting through studies like CCDH's.
- The recurring failure pushes the industry toward hybrid designs that pair community input with algorithmic ranking or paid review, since pure bridging-based consensus appears structurally unable to surface notes on polarized claims.
The trend: Crowd-sourced fact-checking is spreading from X to Meta even as its core defect — most contributed notes never reaching users — remains unsolved, making visibility, not participation, the battleground for whether the model works.