Given the spread of partisan misinformation in politically polarized environments, it is critical to develop interventions that are effective at reducing misinformation sharing in these contexts. Across three online experiments with liberals and conservatives in the U.S. and the UK, we found that crowdsourced accuracy judgments in the form of a Misleading count were effective in reducing participants’ likelihood of sharing misinformation, especially when these judgements reflected an in-group’s opinion (versus general users). The intervention was more effective than other popular interventions, even for misinformation on contested issues (e.g., immigration) and for extreme partisans who highly identify with political leaders. Our work provides new theoretical and practical insights into how to reduce the spread of misinformation in highly polarized contexts.