X is testing a program allowing AI chatbots to create Community Notes

X pilots AI chatbot feature to write Community Notes, fact-checking risky as AIs prone to errors.

: X, formerly Twitter, is testing a feature enabling AI chatbots to create Community Notes, a program aimed at providing context to posts. These AI-generated notes are vetted similarly to human-generated ones, using AI tools like Grok connected via an API. Challenges include AI's tendency to 'hallucinate' and volunteers potentially overwhelmed by AI submissions. This initiative could influence other platforms like Meta, TikTok, and YouTube, which have shown interest in such community-driven fact-checking efforts.

X has launched a pilot program allowing AI-powered “AI Note Writers” to contribute to Community Notes. These bots can generate notes on posts when users request additional context. All AI-written notes are explicitly labeled and must receive enough positive ratings from users with differing viewpoints before they are shown publicly. This ensures the AI contributions meet the same trust standards as human-written notes.

Participation is gated—developers must apply and demonstrate that their bots can produce high-quality content. The bots can lose the privilege to write notes if their outputs are deemed unhelpful. Initially, only a small group of AI bots will be allowed, with the program expected to expand gradually depending on performance and community reception.

Keith Coleman, X’s VP of Product, emphasized that the AI will assist rather than replace humans. The goal is to scale Community Notes without compromising accuracy, especially for posts that attract less attention. Human consensus still determines whether a note is published, maintaining editorial control.

To prevent misinformation, the system includes safeguards against AI hallucinations. Human ratings act as both a gatekeeper and a feedback loop—training the AI via reinforcement learning. Each AI note must pass the same evaluation process used for human-generated notes, ensuring diverse user approval before publication.

The program began testing on July 1, 2025. If successful, it could allow X to scale trustworthy context generation significantly. However, the company acknowledges that trust and transparency are key, so gradual rollout and strict oversight remain central to the initiative.

Sources: The Verge, TechCrunch, Adweek, Tech in Asia, NewsBytes

AIinnovationSocial Media