Bots writing community notes? X thinks it’s a good idea – Quartz
Andrey Rudakov/Bloomberg via Getty Images
Elon Musk’s X is gearing up to let AI chatbots take a stab at fact-checking, piloting a new feature that allows bots to generate Community Notes.
Community Notes, a Twitter-era invention that Musk expanded after buying the platform, lets select users add context to posts — whether it’s clarifying a misleading political claim or pointing out that a viral AI-generated video is, in fact, not real life. Notes only appear after they achieve consensus among users who’ve historically disagreed on past ratings, a system designed to ensure balance and accuracy.
Now, X wants AI in on the action. Keith Coleman, the product executive overseeing Community Notes, told Bloomberg that developers will soon be able to submit their AI agents for approval. The bots will write practice notes behind the scenes, and if they’re deemed helpful (and presumably not hallucinating wildly), they’ll be promoted to public fact-checker.
“They can help deliver a lot more notes faster with less work, but ultimately the decision on what’s helpful enough to show still comes down to humans,” Coleman said. “So we think that combination is incredibly powerful.”
Translation: the bots will churn it out, but humans will still be left holding the final ‘post’ button.
The AI agents can use X’s own Grok chatbot or any other large language model connected via API.
AI’s track record for getting facts right is also — spotty. Models often “hallucinate,” confidently making up information that sounds accurate but isn’t. Even with human checks, there are concerns that an influx of AI-generated notes will overwhelm volunteer reviewers, making them less effective at catching mistakes.
There’s also the risk that AI bots will prioritize sounding polite and helpful over actually correcting misinformation. Recent issues with ChatGPT being overly sycophantic illustrate that, yes, bots want to please you — facts optional.
For now, users won’t see AI-generated Community Notes just yet. X plans to test the bots quietly behind the scenes for a few weeks before deciding whether to unleash them onto the timeline.
