Stop pushing junk science! User corrections of misinformation in science are common and sophisticated.

Misinformation is often equated with blatant falsehoods from dubious sources, overlooking that scientific sources, too, can accidentally publish inaccurate information to large online audiences. This article explores how social media users share, evaluate and address misinformation in science, defined as unreliable research rejected by relevant experts for violating scientific integrity. Using a mixed-methods analysis of Twitter discussions on retracted COVID-19 studies (33,730 tweets and 13,743 replies), the study found that, while most users accepted retracted science at face value, they were also disputed often, with user corrections accounting for nearly half of all replies. Corrections also closely mirrored established experimental recommendations, being mostly elaborative, non-confrontational, and commonly employed one or more debunking strategies. In contrast, polarising and overly convoluted language was rare. The results highlight the significant role that ordinary users play in countering unreliable science 'in the wild' and the need for researchers and policymakers to strengthen support for bottom-up corrective practices.