AI Literacy and Deepfake Detection Competence in Peruvian Higher Education

University students' ability to identify deepfakes and AI-generated content has become educationally relevant, yet the mechanisms through which AI literacy translates into detection competence are not well understood outside North American and European contexts. This study investigated those mechanisms in a sample of 7,259 undergraduate students from 18 Peruvian universities using a mediation and moderation model. PROCESS Model 4 and Model 58 were applied with bias-corrected bootstrapping (5,000 resamples). AI literacy had a significant direct effect on deepfake identification ability (β = 0.524, p < .001). Deepfake awareness (indirect effect = 0.140, 95% CI [0.113, 0.168]) and verification strategies (indirect effect = 0.075, 95% CI [0.056, 0.095]) both mediated this relationship significantly. Prior AI exposure strengthened detection performance, while high digital media consumption without critical training weakened it (β = −0.089, p < .001). The model accounted for 51.0% of the variance in identification ability. AI literacy alone does not produce reliable detection competence; structured training in synthetic content awareness and active verification habits is required, and these findings carry direct implications for curriculum design in Peruvian and broader Latin American higher education.