In two studies we assessed whether simple surface-level textual features of fictional works predict their appeal to readers. Specifically, we focused on the potential role of emotional valence and arousal, the presence of affective themes (i.e., sex, religion, and death), and word concreteness and frequency. In Study 1, we compiled a corpus of 1,471 short stories posted to a creative writing website along with ratings of these stories by readers. Using word lexicons we quantified the textual features of these stories, and then examined whether any of these features predicted ratings using correlation and mixed-effects regression. The presence of abstract, negatively valenced, and commonly used words predicted higher story ratings, with sexual content having a negative impact on ratings. In Study 2, we extended this investigation by examining how textual features relate to appeal for one central aspect of fiction: its characters. A sample of 207 undergraduates wrote brief character sketches, and another group of students rated the depicted characters based on interest, likability, and complexity (n = 144). As revealed in correlations and mixed-effects regressions, characters described with abstract, negative, and arousing words tended to be rated as more interesting and complex. In contrast, characters described with positive and calm words tended to be more likable. Our results succeed in tying specific surface-level textual features to the appeal of stories and characters, highlighting the importance of affect words and abstractness.