Humans spontaneously attribute a wide range of traits to conspecifics based on their facial appearance. Unsurprisingly, previous findings indicate that this ‘person evaluation’ is affected by information provided about the target’s past actions and behaviours. Strikingly, many news items shared on social media sites (e.g., Twitter) describe the actions of individuals who are often shown in accompanying images. This kind of material closely resembles that encountered by participants in previous studies of face-trait learning. We therefore sought to determine whether Twitter posts that pair facial images with favourable and unfavourable biographical information also modulate subsequent trait evaluation of the people depicted. We also assessed whether the effects of this information-valence manipulation were attenuated by the presence of the “disputed tag”, introduced by Twitter as a means to combat the influence of fake-news. Across two preregistered experiments, we found that fictional tweets that paired facial images with details of the person’s positive or negative actions affected the extent to which readers subsequently judged the faces depicted to be trustworthy. When the rating phase followed immediately after the study phase, the presence of the disputed tag attenuated the effect of the behavioural information (Experiment 1: N = 128; Mage = 34.06; 89 female, 36 male, 3 non-binary; 116 White British). However, when the rating phase was conducted after a 10-minute delay, the presence of the disputed tag had no significant effect (Experiment 2: N = 128; Mage = 29.12; 78 female, 44 male, 4 non-binary, 2 prefer not to say; 110 White British). Our findings suggest that disputed tags may have relatively little impact on the long-term face-trait learning that occurs via social media. As such, fake news stories may have considerable potential to shape users’ person evaluation.