In everyday social interactions, people’s facial expressions sometimes reflect genuine emotion (e.g., anger in response to a misbehaving child) and sometimes do not (e.g., smiling for a school photo). There is increasing theoretical interest in this distinction, but little is known about perceived emotion genuineness for existing facial expression databases. We present a new method for rating perceived genuineness using a neutral-midpoint scale (–7 = completely fake; 0 = don’t know; +7 = completely genuine) that, unlike previous methods, provides data on both relative and absolute perceptions. Normative ratings from typically developing adults for five emotions (anger, disgust, fear, sadness, and happiness) provide three key contributions. First, the widely used Pictures of Facial Affect (PoFA; i.e., “the Ekman faces”) and the Radboud Faces Database (RaFD) are typically perceived as not showing genuine emotion. Also, in the only published set for which the actual emotional states of the displayers are known (via self-report; the McLellan faces), percepts of emotion genuineness often do not match actual emotion genuineness. Second, we provide genuine/fake norms for 558 faces from several sources (PoFA, RaFD, KDEF, Gur, FacePlace, McLellan, News media), including a list of 143 stimuli that are event-elicited (rather than posed) and, congruently, perceived as reflecting genuine emotion. Third, using the norms we develop sets of perceived-as-genuine (from event-elicited sources) and perceived-as-fake (from posed sources) stimuli, matched on sex, viewpoint, eye-gaze direction, and rated intensity. We also outline the many types of research questions that these norms and stimulus sets could be used to answer.