Fake news can foster political polarization, foment division between groups, and encourage malicious behavior. Misinformation has cast doubt on the integrity of democratic elections, downplayed the serious- ness of COVID-19, and increased vaccine hesitancy. Given the leading role that online groups play in the dissemination of fake news, in this research we examined how group-level factors contribute to sharing mis- information. By unobtrusively tracking interactions among 51,537 Twitter user dyads longitudinally over two time periods (n = 103,074), we found that group members who did not conform to the behavior of other group members by sharing fake news were subjected to reduced social interaction over time. We aug- mented this unique, ecologically valid behavioral data with another digital field study (N = 178,411) and five experiments to disentangle some of the causal mechanisms driving the observed effects. We found that social costs were higher for not sharing fake news versus other content, that specific types of deviant group members faced the greatest social costs, and that social costs explained fake news sharing above and beyond partisan identity and subjective accuracy assessments. Overall, our work illuminates the role of conformity pressure as a critical antecedent of the spread of misinformation.