This work seeks to address educational technologies through critical questions, avoiding the idea of neutrality that is commonly attributed to them, including in academic literature. The general objective is to analyze discourses on artificial intelligence (AI) promoted by educational technology companies (edtechs) that offer technologies aimed at teaching-learning. The specific objectives are: 1) to examine how AI is conceived in edtech discourses; 2) to investigate the roles assigned by edtechs to technologies in teaching and learning processes; and 3) to characterize the conceptions of education conveyed by these discourses. Critical Discourse Analysis was the theoretical-methodological framework that guided the analyses, starting from the categories of assumptions, lexical choices and modality. The corpus is composed of texts taken from websites and seventeen Instagram posts from three selected companies: Letrus, Educacross and Jovens Gênios. In the material analyzed, AI is conceived as a solution to various educational problems and positioned either as a subject or as a tool for learning. AI appears with the role of personalizing education, making learning more meaningful and education more objective and efficient, based on data, democratic and innovative, in addition to being able to overcome deficiencies in teaching. Companies disseminate a conception of education based on performance, organized by skills and abilities, markedly individualistic, marginalizing the relationship between educator and student. The discourses express an ideal of education that values individual qualification to the detriment of the dimensions of socialization and subjectivation, therefore far from ideals of broad training and social transformation.