CANote AI system helps lay users write expert-level fact-checking notes

A new research paper presents CANote, an AI system designed to reduce the burden of writing high-quality fact-checking notes on platforms like X's Community Notes. The tool scaffolds the entire workflow: it first extracts subclaims from a social media post, then retrieves and explicitly links relevant evidence to each subclaim (provenance), and finally generates a neutral, structural draft that users can refine. This approach aims to support human reasoning while ensuring evidence-based debunking.

In a controlled experiment with 52 professional fact-checkers and 52 lay users, CANote produced notes with significantly higher quality scores compared to manual writing. Crucially, lay users using CANote achieved note quality comparable to experts writing manually. Task completion time and perceived cognitive load remained similar to manual drafting, but user satisfaction increased. However, the assistance introduced a trade-off: users reported a reduced sense of ownership and control over the final note. The findings suggest that carefully designed AI scaffolds can democratize fact-checking without overwhelming contributors, though maintaining user agency remains a challenge.