X rolls out AI tools to curb manipulation of community notes

X, formerly known as Twitter, has quietly rolled out an update to its Community Notes feature aimed at detecting and curbing coordinated manipulation by contributors. The company says the new system introduces safeguards against users working together to unduly influence the visibility of fact-checking notes.

According to a statement issued by the platform, the update expands the existing detection mechanism to cover both note writers and raters, automatically neutralising the effect of suspiciously correlated ratings. “When the scoring algorithm detects anomalous correlations in ratings, it automatically prevents those ratings from contributing to notes achieving helpful status,” X explained.

The move comes in response to growing concerns about the integrity of the Community Notes system. Analysts and watchdog groups have raised alarms over organised campaigns by contributors attempting to suppress notes they disagree with ideologically. Studies have shown that up to 85% of Community Notes are never shown to users, often due to contributors’ failure to reach consensus, especially on politically polarising topics.

Under X’s current model, a note only appears if users from opposing ideological viewpoints agree on its value. This has led to the system struggling to moderate highly contentious discussions around issues like voter fraud, gender identity, abortion rights, and U.S. foreign policy. As a result, misleading or controversial content often remains unflagged on the platform.

Despite these concerns, X has committed to improving Community Notes, having introduced multiple technical changes over the past year. Whether these adjustments will restore confidence in the system, or serve as a model for other platforms exploring community-led moderation—remains to be seen.

The company did not specify how it would ensure transparency or accountability in its latest detection method.