X, the social platform formerly known as Twitter, is experimenting with a new feature that will allow AI chatbots to generate Community Notes. This fact-checking system, which lets users add context to misleading or incomplete posts, has become a core part of X under Elon Musk’s ownership. The idea behind Community Notes is that once enough users with differing perspectives agree on the same note, it becomes visible to the public. Now, X is testing whether AI models can contribute to this process.
The AI-generated notes can come from X’s chatbot, Grok, or from external AI tools integrated through an API. These notes will be treated just like those submitted by human contributors; they must still pass the same community vetting process before being published. X says the aim is not to replace human judgment but to build a system where AI helps generate useful notes, which are then reviewed by people before going live. The company’s research team suggests that human feedback should guide AI outputs, creating a loop where both work together to produce better context.
Still, the use of AI in fact-checking raises concerns. Language models are known to “hallucinate,” meaning they sometimes generate inaccurate or misleading information while sounding confident. If a chatbot prioritises sounding helpful over being accurate, its contributions may backfire, especially when used to fact-check sensitive or political content. There’s also the question of scale. If AI-generated notes flood the system, volunteer human reviewers might feel overwhelmed or lose motivation to keep up.
The challenge will be maintaining accuracy without over-relying on automation. Even with human checks in place, third-party models embedded into X’s system could behave unpredictably. Some, like OpenAI’s ChatGPT, have had issues with overly agreeable or factually flawed responses. X will need to make sure that the AI agents it uses or allows others to use are both transparent and reliable.
For now, the AI-generated Community Notes feature will be tested on a small scale. If the trial shows that AI can assist without undermining trust in the system, X may roll it out more broadly. Until then, the platform is treating this as a controlled experiment, one that blends human judgment with the speed and reach of artificial intelligence.