TAIL: A Test of AI Literacy for Adolescents and Teachers

Most instruments that claim to measure AI literacy inquire how competent people feel rather than test what they know, and almost none has been validated for both students and teachers. We developed the Test of AI Literacy (TAIL), a multiple-choice test of AI literacy built on an established four-aspect framework and piloted with cognitive interviews and item analysis. We evaluated it among Czech secondary-school students tested in fall 2025 (n = 2,441) and six months later (n = 1,669; 897 linked across the two measurement waves), and among teachers (n = 564). In every sample the items fitted a one-factor model, whereas the four content aspects could not be distinguished as separate factors. Three weakly discriminating items were removed after the first wave, leaving a 17-item form with internal consistency of .79 to .87 and item response theory reliability of .80 to .84. Item parameters were almost identical across waves. Six-month test-retest reliability was .63, with a standard error of measurement of 2.4 points. Teachers scored .41 standard deviations above students, and teachers of information and communication technologies, grammar-school teachers, and students with a university-educated parent scored higher, as predicted. Teachers' self-rated AI self-efficacy was practically unrelated to their tested knowledge, and students who used and relied on AI more knew less about it. TAIL is a short, knowledge-based measure that places adolescents and their teachers on one scale.