X tests 'Under the Hood' tool to reveal visibility-limiting labels

X is testing a new transparency feature called "Under the Hood" that allows users to determine if the platform's ranking algorithm has limited the visibility of their posts. The tool aims to provide direct insight into the mechanisms that have long been the subject of "shadowbanning" accusations.

Currently available to a small group of testers, the tool is restricted to accounts at least one year old with a minimum of 10 posts in the last month. Eligible users can access a new "Under the Hood" page within the app's settings to download a JSON file. Rather than providing data on a per-post basis, the file contains aggregate statistics on visibility-limiting labels applied to the account and its content over the previous calendar month. These labels include categories for spam, NSFW content, violence, impersonation, hateful conduct, and a specific "civic integrity" flag.

Algorithmic Open-Sourcing

This feature is part of a broader strategic shift toward transparency under the leadership of Elon Musk. X has open-sourced its "For You" timeline ranking algorithm on GitHub under the Apache v2 license, allowing external developers and researchers to inspect the code that governs content distribution. This follows years of controversy where political figures and users accused the platform of limiting reach without notification—a practice the company previously denied.

Keith Coleman, X VP of Product, described the move as "an unprecedented level of transparency into the X algorithm." Coleman stated that the company's goal is for the public to assess how posts are distributed, vet the platform as a "level playing field," and provide critiques to help the company improve the system.

Implications for Moderation

By providing the actual labels used to restrict reach and the underlying code, X is attempting to shift the debate over shadowbanning from anecdotal complaints to data-driven audits. This represents a significant departure from the opaque moderation policies typical of major social media platforms, where visibility limits are usually hidden from the user.

However, the transparency is not absolute. X continues to withhold certain portions of its codebase, specifically logic designed to prevent "gaming" of the system and code related to advertising. This suggests a balance between public accountability and the need to protect the platform's commercial interests and security.

What to Watch

As the test expands, the industry will be watching to see if this data actually resolves user grievances or if the aggregate nature of the JSON files proves too vague for meaningful audits. It remains to be seen if other social media giants will face pressure to adopt similar verifiable transparency tools for their own proprietary ranking algorithms.