In our project, we investigate the scientific validity of a specific self-built Adaptive Learning Tool in the field of dynamic geometry with a focus on the individual learning pathways of a highly diverse student population. A total of 164 children in Luxembourgish elementary schools, aged between 10 and 13 years, acted as the test group and explored elementary geometric concepts in a sequence of learning assignments created with GeoGebra, a dynamic mathematics system which is integrated into the computer-assisted testing framework TAO. They actively built new knowledge in an autonomous way and at their own pace with only minor support interventions by their teacher. Based on easily exploitable data collected within a sequence of exploratory learning assignments, the GEOGEBRATAO tool analyses the answers provided by the child and performs a diagnostic of the child’s competencies in geometry. With respect to this outcome, the tool manages to identify children struggling with geometry concepts and subsequently proposes a differentiated individual pathway through scaffolding and feedback practices. The children can voluntarily watch short video clips aimed to help them better understand any task that they might have difficulty with. A spaced repetition feature is another highly useful component of the tool. Pre- and post-test results show that the test group (working with the GEOGEBRATAO tool) and a parallel working control group (following a traditional paper-and-pencil geometry course), increased their geometry skills and knowledge through the training program, with the test group performing even better in items related to dynamic geometry. In addition, a more precise analysis within clusters based on similar performances in both pre- and post-tests and the child’s progress within the GEOGEBRATAO activities, provides evidence of some common ways of working with our educational technology tool, leading to overall improvement at an individualized level.