This article outlines the main results and methodological challenges of a large scale survey on actual digital skills. A test covering three main dimensions of digital literacy (theoretical, operational and evaluation skills) was administered to a random sample of 65 third-year high school classes, producing data on 980 students. Items include knowledge questions, situation-based questions and tasks to be performed online. A Rasch-type model was used to score the results. In agreement with literature, the sample performed better in operational skills, whilst showing a particularly poor performance regarding evaluation skills (although for this dimension the test shows reliability issues). Through a robust regression analysis we investigate if a skills divide based on ascriptive differences, gender and family cultural background, exists among the students. It emerges that cultural background has a significant effect, which is stronger on operational skills, while gender shows a more definite impact on theoretical knowledge. Methodological problems related to the measurement process are discussed and it is pointed out that a lack of standardised criteria to interpret substantive score differences exists in this field of study.