his study investigated the added value of real-time adaptive feedback on seventh graders' performances in tablet-based geometry learning. To isolate the effects of the medium (ie, tablet) from those of the feedback, three groups were compared: paper-and-pencil, pen-based tablet without feedback and pen-based tablet with feedback. The feedback was provided by a tutoring system based on an artificial intelligence that automatically interpreted students' pen strokes on the screen. A total of 85 French students drew three geometric shapes, either on paper or on a tablet, and then performed a transfer task on paper. Results showed that using a tablet without feedback did not improve learning but seemed to enhance interest in the task compared to the paper-and-pencil group. Students in the tablet with feedback group performed significantly better than the other two groups on learning, as well as on transfer. This study was the first to combine media comparison and added-value approaches to test the effects on students' geometry performances of using a new educational app on a pen-based tablet in a naturalistic classroom environment. Results showed that it was not the medium used but the intelligent tutoring system-based feedback that improved students' performance. Our data therefore indicate that artificial intelligence is a promising way of providing learners with real-time adaptive feedback in order to improve their performances.his study investigated the added value of real-time adaptive feedback on seventh graders' performances in tablet-based geometry learning. To isolate the effects of the medium (ie, tablet) from those of the feedback, three groups were compared: paper-and-pencil, pen-based tablet without feedback and pen-based tablet with feedback. The feedback was provided by a tutoring system based on an artificial intelligence that automatically interpreted students' pen strokes on the screen. A total of 85 French students drew three geometric shapes, either on paper or on a tablet, and then performed a transfer task on paper. Results showed that using a tablet without feedback did not improve learning but seemed to enhance interest in the task compared to the paper-and-pencil group. Students in the tablet with feedback group performed significantly better than the other two groups on learning, as well as on transfer. This study was the first to combine media comparison and added-value approaches to test the effects on students' geometry performances of using a new educational app on a pen-based tablet in a naturalistic classroom environment. Results showed that it was not the medium used but the intelligent tutoring system-based feedback that improved students' performance. Our data therefore indicate that artificial intelligence is a promising way of providing learners with real-time adaptive feedback in order to improve their performances.