Artificial intelligence chatbots are known for providing suspicious information and hallucinations regularly, making it a terrible prospect for fact-checked characters. However, Elon Musk’s X (néeTwitter) plans to deploy AI agents to help fill the gap in the infamous slow-responsive community notes, which will appear this month. What might be wrong?
According to the public announcement, the new model will allow developers to submit AI agents to review AI agents. Agents will test behind the scenes to write practice notes to see their performance. If they provide useful information, they will get a green light, and can stream live on the platform and write notes. These notes still have to be approved by human reviewers on community notes and still need to be found useful by people of various perspectives (the way the metric is determined is somewhat opaque).
According to Bloomberg, developers who submit their own agents can be powered by AI models, so despite having direct connections with Musk, users won’t be locked in Grok (perhaps because Musk simply can’t stop Grok from being awakened, no matter how hard he tries). The company’s expectation is that AI-generated notes will significantly increase the number of notes published on the platform.
They kind of need for AI, because human-generated notes have been reportedly dropped from the cliff. NBC News reported last month found that the number of notes posted on the platform cuts from 120,000 in January to 60,000 in May 2025. Basically, participation with fact-checking services collapsed.
There may be many factors. First of all, the platform is a bit like a shit show. A Bloomberg analysis found that it basically took about 14 hours to attach a note to a post with fake or misleading information after its main virus cycle passed. Differences among community notes contributors also resulted in fact check failures and about a quarter of them were published due to disagreement among the assessors. This number gets higher when it comes to positively controversial issues such as Russia’s invasion of Ukraine, which ends up being removed over 40% of the notes.
Then there is a guy who owns the site, despite actively promoting community notes as a huge solution to misinformation, spending more and more time on it. Musk claimed earlier this year that there was no evidence that government actors and legacy media could direct community notes, instilling distrust throughout the process. Do you know what won’t make the system harder to play? Release the robot on the issue.
