X’s Open-Source Algorithm Shows AI-Content Downranking
cnBeta, citing X’s open-source recommendation algorithm and transparency tools, reports that X can classify suspected AI-generated posts and accounts. The account-level “llm_slop_user” label carries a seven-day SpamHighRecall designation and suppresses recommendations to non-followers, while “llm_slop_post” receives a 30-day RiskyHighVizReply label and suppresses reply and mention notifications to users who do not follow the author. X’s transparency tools say existing followers can still see affected posts, and cnBeta reports that the labels were already present in the repository by September 4.
The report provides concrete details about how X’s recommendation system may reduce the distribution of suspected AI-generated content while leaving content visible to followers.