While current research on AI-generated content detection predominantly focuses on algorithmic solutions, limited attention has been given to end-users’ ability to discern AI-generated from human-produced material. This study investigates Polish undergraduate students’ self-assessed AI literacy, specifically their perceived ability to recognize AI-generated text, images, and videos. A quantitative survey (n = 350) comprising 32 Likert-scale items examined students’ engagement with AI tools, exposure to AI-generated media, perceived recognition capabilities, and online information evaluation practices. Over 80% reported encounters with AI-generated content and over 70% expressed confidence in identifying AI-generated videos. Nonetheless, the infrequent use of verification tools indicates potential overestimation of their competencies. This discrepancy may reflect cognitive biases, as existing literature underscores the difficulty of such distinctions. A majority of participants supported mandatory labeling of AI-generated content. These findings highlight a critical need for empirical validation of users’ detection abilities and underscore the importance of developing robust AI literacy frameworks.