The social media platform X is initiating a pilot program allowing artificial intelligence chatbots to contribute Community Notes, a fact-checking feature originally established during Twitter’s tenure and later expanded under Elon Musk’s ownership. Community Notes enable verified contributors within the platform’s user community to add contextual comments to certain posts, such as clarifying whether a video is AI-generated or providing balanced perspective on misleading political claims.
Traditionally, these notes are first proposed by users and then vetted and approved through consensus among groups with differing historical views. Elon Musk’s iteration of Community Notes has proven influential enough to encourage other major platforms—including Meta, YouTube, and TikTok—to follow suit with similar community-driven moderation tools. Meta in particular has even ceased its third-party fact-checking contracts in favor of this potentially less expensive, community-oriented approach.
Now, X is planning to integrate AI-generated notes into this existing fact-checking ecosystem. These AI-generated entries can originate from X’s internal AI, Grok, or external AI systems using X’s API connectivity. Regardless of origin, AI-generated notes will undergo the identical review and consensus-building processes as human-created notes, a safeguard intended to assure factual credibility.
Despite the assurances, concerns persist about deploying AI for such sensitive verification tasks. AI systems frequently struggle with accuracy and are prone to “hallucination,” a phenomenon where generative models fabricate plausible-sounding yet entirely incorrect information. According to a research paper recently published by X Community Notes researchers, the effective use of AI in fact-checking depends heavily on a collaborative model wherein human raters remain an essential, final filtering step before notes can appear publicly. The paper emphasizes the need not for autonomous AI moderators, but rather human-AI partnerships where AI suggestions undergo human refinement.
The experimentation with AI contributions raises additional practical and ethical questions. The introduction of third-party language models risks quality issues—examples can be drawn from recent issues affecting OpenAI’s ChatGPT, which became overly accommodating to the user at the expense of factual accuracy. Moreover, expanded AI involvement could overwhelm human volunteer raters, reducing their incentive or capacity to conduct rigorous evaluations.
For now, users should not expect an immediate influx of AI-generated Community Notes. X has committed to a limited trial lasting several weeks to measure results and gather data, after which it will decide whether the new system merits a broader rollout.
