If you're a regular poster on X (formerly Twitter), you might assume that accumulating likes is the clearest indicator of your content's success. However, X has recently unveiled key insights into its recommendation algorithm, codenamed Phoenix, revealing that likes are far from the most influential factor in determining a post's visibility and virality.
Understanding X's Phoenix Algorithm
X has made parts of its recommendation algorithm's source code public, providing creators with a clearer picture of how posts are selected and ranked for the coveted “For You” feed. At its core, Phoenix predicts a user's likelihood of interacting with a post and uses these predictions to calculate its ranking score.
What Truly Boosts X Post Visibility?
The data from X's Phoenix system highlights a significant difference in the weight assigned to various engagement types. Surprisingly, simply collecting likes may not be enough for greater visibility. Here's a breakdown of how different interactions are weighted compared to a 'like' (which has a base positive score of 0.5):
- URL Shares: If Phoenix predicts you will share a post by copying its URL, this action carries an astonishing 80 times the weight of a single like. This translates to roughly 40 times the positive score of a like.
- Replies, Quote Posts, Direct Message Shares: Each of these interactions is assigned a weight of 5.
- Follows: A predicted follow carries a weight of 8.
- Reposts: A repost has a weight of 1.
These figures clearly indicate that active engagement, particularly actions that spread content beyond the immediate follower base, has a much stronger influence on a post's ranking score than passive approval through likes.
Negative Reactions Significantly Reduce Reach
Beyond positive interactions, Phoenix also heavily factors in actions that signal users do not want to see certain content. These negative signals can have a disproportionately large impact on a post's ranking, drastically reducing its visibility:
- Predicted Report: Carries a severe negative weight of -234.
- Mute: Weighted around -59.
- “Not Interested”: Carries a negative weight of -43.
- Block: Assigned a negative weight of -31.
Compared to the 0.5 weight of a like, these negative signals demonstrate X's strong emphasis on user experience and content moderation, allowing users to actively shape their feeds by signaling unwanted content.
X Tests 'Under the Hood' Feature
In a move towards greater transparency, X has also introduced a pilot program called 'Under the Hood'. This feature allows a randomised group of selected users to view aggregate information about labels that may affect the visibility of their posts or accounts. The wider availability of 'Under the Hood' will depend on user feedback from this initial pilot phase.
For content creators, the takeaway is clear: while likes offer a basic form of validation, X's algorithm considers a much broader and deeper set of signals when deciding what appears in users' “For You” feeds. Focusing on content that encourages shares, replies, and follows will be far more effective in achieving widespread visibility and virality on the platform.