Although perceptions of subjective well-being (SWB) in unacquainted others have been shown to play a major role in impression formation, little is known about how accurate such perceptions are. In two original studies and one pre-registered replication, we explored the accuracy of life satisfaction and happiness judgments from texts and its underlying mechanisms (use of linguistic cues). Participants filled in life satisfaction and happiness measures and completed a brief writing task. Another sample of participants judged the targets’ life satisfaction and happiness from the obtained texts. All three studies demonstrated a small to moderate self-other agreement. A linguistic analysis showed that targets with higher (vs. lower) scores on SWB were less likely to use negation words in their texts, which allowed observers to make accurate judgment of their SWB level. Two studies pointed at negative emotion words as valid and positive emotion words as invalid (but often used) cues to happiness, yet these effects did not replicate in Study 3.