The purpose of this text is to propose some methodological assumptions for the control of administrative decisions taken with artificial intelligence. Although the use of this type of tools in the exercise of public powers poses new challenges, this should not be understood as a field that escapes from the classical elements of public control. On the contrary, a dynamic interpretation of the criteria and perspectives traditionally used in the control of these powers constitute inputs capable of offering some answers to these questions (or at least a first approximation to them). We then propose a methodology focused on the analysis of the impact that artificial intelligence could have on the legality of the different elements addressed in the “general theory of the administrative act”. In this way, the intellectual decomposition of formal administrative activity in terms of its essential elements contributes not only to a better understanding of the impact of artificial intelligence on such activity, but also to the very definition of its legality.