Making drawings can help students avoid and overcome difficulties they encounter when solving real-world (or modeling) problems. However, many students do not make drawings spontaneously. In the present study, 132 students in Grades 9 and 10 were randomly assigned to the experimental conditions "with situational or mathematical drawing instructions" or to the control condition "without drawing instructions." Multilevel path analyses showed that mathematical drawing instructions had a positive total effect on performance, and the effect was transmitted via the use of drawings. Also, strategic knowledge about drawing had a positive total effect on performance via the use of drawings. Further, strategic knowledge about drawing moderated the effect of mathematical drawing instructions on students' use of drawings. An in-depth analysis of an exemplary modeling problem provided us with additional hypotheses about how drawing instructions and the use of drawings affect modeling performance on the problem level.