Tech

X opens source ranking algorithm to allow public audit of shadowbanning claims

The move expands the open-source repository by up to 15 times, enabling external developers to run ranking systems and users to check if their content has been deprioritised by the platform’s algorithms.

Editorial persona
Owen Mercer
Markets and Finance Editor
Published
Draft
Source: TechCrunch · View original source
X open sources its ranking algorithm, letting users see if they’ve been ‘shadowbanned’
Social media giant releases core codebase on GitHub and launches transparency tools for users

X has significantly expanded its open-source codebase, releasing the source code for its 'For You' feed and core ranking engine on GitHub under the Apache v2 license. The company is also launching transparency tools that enable users to determine if their accounts or posts have been impacted by ranking systems. A new feature allows users with ten or more posts in the past month to download aggregate statistics as a JSON file, revealing any labels applied to their content. This initiative aims to address concerns regarding algorithmic influence on politics and misinformation, allowing for public auditing and critique of the platform's distribution systems.

The codebase is now roughly 10 to 15 times larger than before, including model configuration, filter, and core ranking system details. Users can run systems like the ranker and score outside of X. Non-technical users can use an LLM to interpret the downloaded JSON data by pointing it at X’s GitHub repo. The transparency tool is initially rolling out to a test group of accounts at least a year old. External researchers were able to get the 'score' running outside of X ahead of the launch. Developers can submit pull requests to the GitHub repository, which X engineers will consider. Systems using Grok to predict rule violations are excluded from the open-source release to prevent bad actors from gaming the system.

It is unclear how many accounts are in the initial test group for the transparency tool. The extent to which submitted pull requests will actually be incorporated into the algorithm is not specified, only that they will be 'considered'. The specific impact of the open-sourced code on reducing misinformation or political bias is not yet measurable.

X is owned by a trillionaire who helped President Trump get elected. Twitter faced allegations from Republicans in Congress in earlier years that it leaned too left and 'shadowbanned' posts, which the company denied. X has become less transparent overall under Musk after becoming a private company again, no longer required to report to the SEC. X is less forthcoming with user metrics, growth, revenue, or government takedown requests. X has a crowdsourced fact-checking system called Community Notes.

The release represents a shift in how the platform handles scrutiny over its distribution mechanisms. By making the parameters that weight different signals public, X aims to allow users and researchers to assess whether the playing field is level. This comes amid broader debates about the role of algorithms in shaping political discourse and the spread of information online. The platform has historically faced accusations of bias, with past congressional inquiries focusing on whether certain viewpoints were suppressed.

While the open-source initiative allows for external verification of ranking logic, it does not extend to all aspects of the platform’s moderation infrastructure. Systems utilising Grok to predict potential rule violations remain proprietary, a decision cited as necessary to prevent bad actors from exploiting the code to circumvent safety measures. This selective transparency highlights the tension between openness and the need to maintain platform integrity against coordinated manipulation.

The move to open-source the ranking engine also invites a new form of community engagement with the platform’s technical architecture. Developers can now submit pull requests to the repository, offering a potential avenue for external contributions to the algorithm’s development. Although not all submissions will be integrated, the possibility of public code improvement marks a departure from the traditional closed-source model typical of major social media platforms. This approach could foster greater trust among users who have long questioned the opacity of content distribution systems.

Continue reading

More from Tech

Read next: Apple’s Screen Time offers tools to rein in iPhone usage
Read next: GameCube’s library still commands attention 25 years on
Read next: US AI leaders urge restraint as Trump team prioritises China competition