Deepfake videos are hyper-realistic manipulated videos, which have now become rather challenging to distinguish from genuine ones owing to insurmountable technology-related advances. While the majority of research is focused on algorithmic approaches to deepfake detection, significantly less attention has been given to our understanding of human capabilities toward recognizing them. Thus, the objective of the current study is to identify critical factors like social media usage, age, familiarity with the subject, multimodal factors influencing human performance in recognizing a deepfake. The findings suggest that individuals detect deepfake videos containing familiar faces better than those able to detect deepfakes of unfamiliar faces. It was further revealed that persons who frequently use social media are better equipped to spot deepfakes. While performance in detection declines with age, gender does not appear to affect deepfake recognition. This study also shows a comparison of detection accuracy in still images, silent videos, and videos with audio.