News

‎X Reveals How Its Algorithm Decides Which Posts Go Viral

X has released the source code behind its recommendation algorithm, offering users a clearer look at how posts are selected for the platform’s “For Yo...

By Wavers Multimedia

‎X Reveals How Its Algorithm Decides Which Posts Go Viral
X has released the source code behind its recommendation algorithm, offering users a clearer look at how posts are selected for the platform’s “For You” feed.
‎
‎The algorithm, known as Phoenix, ranks posts according to how likely they are to generate engagement. The newly released details show that different types of interactions carry different levels of weight.
‎
‎Shares through direct messages, quotes, replies and URL sharing can have a stronger effect on a post’s distribution than likes. In particular, URL shares receive a significantly higher weighting than likes in the recommendation system.
‎
‎The system also takes negative interactions into account. Actions such as reporting, muting, blocking or selecting “not interested” can reduce the visibility of a post.
‎
‎That means content designed purely to provoke users may not necessarily gain the reach its creator expects. Negative feedback can work against a post as the system evaluates whether it should appear to more people.
‎
‎X has also introduced a pilot feature called “Under the Hood”, which allows selected users to see how certain content labels, including spam and adult-content classifications, can affect the visibility of their posts.
‎
‎The company says the move is part of its broader effort to make the recommendation system more transparent and give users a better understanding of how content reaches the “For You” timeline.
‎
‎The release gives creators a clearer picture of what can influence distribution on X, showing that engagement quality and user reactions matter alongside the number of likes a post receives.
‎
‎The platform is expected to continue refining the system based on feedback as it works toward greater transparency around how recommendations are made.