The use of evidence-based practices can be regarded as the gold standard in technology-based learning and instruction. A steadily adopted educational practice is technology-mediated non-interactive teaching, in which students generate explanations of the previously learned contents to a fictitious audience by means of technologies (e.g., video, messenger). Although recent laboratory studies documented benefits of non-interactive teaching, field-oriented evidence is scarce. Research is needed to examine how laboratory evidence applies to authentic learning environments with school students and to determine whether these effects are generalizable to different authentic contexts. We applied a ManyClasses study to a) examine the generalizability of technology-based non-interactive teaching and b) explore context-related (domain, school type), demographical-related (age, gender, language), and implementation-related (grading, medium, timing) boundary conditions. In collaboration with teachers, we realized k = 20 different teaching units (each consisting of two lessons) in authentic settings across various school types and domains. Using a within-participants design, school students (N = 191) either taught the previously learned contents by means of technology to a fictitious peer or retrieved the contents in mind after the lesson. Results showed no main effect of non-interactive teaching; but domain and school type moderated the learning activity. The findings indicate that non-interactive teaching is not effective per se, but rather depends on the instructional contexts in which it is implemented. The investigation of the teaching effect with new approaches allows, for the first time, more generalizable conclusions to be drawn about non-interactive teaching with technology for students in authentic settings.