Abstract. The use of emojis in digital communication has become increasingly popular, but how emojis are processed and integrated in reading processes remains underexplored. This study used eye-tracking to monitor university students’ ( n = 47) eye movements while reading single-line text messages with a face emoji embedded medially. Messages contained a semantically congruent emoji (e.g., That’s good news 😊 tell me more), a semantically incongruent emoji (e.g., That’s good news ☹ tell me more), or a dash (e.g., That’s good news – tell me more). Results revealed that emoji congruency did not influence early fixation measures (first fixation duration and gaze duration), nor the probability of regressions. However, there was a significant congruency effect in total reading time and trial dwell time, showing that incongruence incurred a processing cost. The present results extend previously reported semantic congruency effects in sentence reading to the processing of emojis. This result suggests that the semantic content conveyed by face emojis is integrated with sentence context late in processing. We further found that the use of congruent emojis improved the relationship between sender and receiver: Ratings collected separately suggested that message senders were liked better if they included congruent than incongruent emojis. Overall, emojis attracted attention: Participants were twice as likely to fixate on emojis than on dashes, and to fixate on emojis for longer.