Purpose In many contexts in which high‐stakes lies occur (such as security settings or the courtroom), observers must evaluate whether the stories they hear are credible. However, little research has evaluated the ability of observers to detect high‐stakes lies, nor the influence of the manner in which the deception is presented on judgment accuracy. This study investigated whether the presentation modality of high‐stakes lies influences both explicit and implicit deception detection accuracy. Methods Participants (N = 231) were randomly assigned to one of four presentation modalities: audiovisual, video‐only, audio‐only, or transcript‐only and asked to evaluate the honesty of targets – half of whom were sincere and half deceptive killers – making a plea for the return of a missing relative both explicitly (direct lie/truth decision) and implicitly (via emotional reactions). Results Overall, explicit deception detection accuracy was slightly above chance (M = 52.5%), and honest pleas were accurately identified at a higher rate than deceptive pleas. Although there were no differences in overall accuracy across modality, observers reading transcripts exhibited a truth bias, which resulted in them detecting truthful pleas at a higher rate than with the other groups. Although explicit accuracy was at the level of chance, implicit reactions indicated that observers were able to unconsciously discern liars from truth‐tellers. Conclusions Despite the high‐stakes nature of the lies presented here, they were difficult to detect. Lies presented via written language were missed at a higher rate when assessed using explicit but not implicit judgments.