In an online experiment (N = 339), we explored sharing of news headlines on social media. Unbeknownst to participants, half the headlines were real news and the other half fake news. Participants first chose which headlines to share. After a 10-minute retention interval, participants completed an unexpected headline source monitoring test. Participants provided old-new recognition judgments and indicated their remembered behavior (i.e., shared vs. not shared) for each recognized headline. Multinomial processing tree (MPT) analyses revealed little impact of headline truth value on memory parameters. True headlines were shared more often than fake headlines. While recognition memory was better for shared headlines, sharing memory was better for not shared headlines. Hierarchical MPT analyses additionally explored correlations between model parameters and the impact of individual differences in cognitive reflection and open-minded thinking on model parameters. In line with previous findings, results suggest a “social media sharing mindset” characterized by less analytic processing.