X Tests Tool to Reveal "Shadow Banning" and Explain Post Visibility Drops

X is currently testing a new feature designed to increase transparency around how the platform’s algorithms handle user posts. The tool aims to identify restrictions that may affect content visibility and distribution, aligning with the company’s broader effort to disclose more information about its recommendation algorithm.

How the Visibility Tool Works

The feature is available to a limited group of users and can be accessed via the “Advanced Options” page in the app’s settings. Eligible users may download account-related data and search for information on labels that could influence the visibility of their accounts and posts.

This data does not provide a detailed, post-by-post report. Instead, it delivers aggregated statistics covering the previous month — a format that may pose interpretation challenges for some users.

For example, the data indicates whether certain posts received classifications that restrict their reach to non-followers or younger users. Such classifications relate to content deemed unsuitable, violent, or spam-like.

Expanded Algorithm Transparency

The rollout coincides with the release of additional components of X’s ranking algorithm — specifically those powering the “For You” section — as open-source code. X states this move offers users, developers, and researchers broader insight into how the platform determines which content qualifies for inclusion in recommendations.

The newly disclosed classifications include categories tied to inappropriate content, spam, violence, impersonation, and hate speech, alongside other specialized labels.

However, X has not released all details of its recommendation system. The company retains certain information it considers vulnerable to algorithmic manipulation. The current open-source release also excludes advertising-related systems and other platform components.

This development follows earlier statements by Elon Musk affirming plans to make the code governing X’s system open source and subject to external review. The present release does not constitute full implementation of that vision, though additional classifications and data may be added in future updates.