Learner-generated drawings are regarded as a useful visualization strategy for reading comprehension, acquiring knowledge and problem solving. Previous studies in mathematical education have often shown a positive effect of this strategy, especially when working on complex, reality-based tasks. Findings from research on strategies indicate that the use and effect of drawings could depend on the respective mathematical topic. We examined this hypothesis by investigating the effects of the prompt to make a drawing on modelling competence regarding the mathematical topics Pythagorean theorem and Linear Functions in a study with ninth-grade learners (n = 169). Quantitative analyses show that (1) the frequency and quality of sketches are significantly higher in the Pythagorean theorem than in the topic Linear Functions, even when drawings are prompted; (2) there is a high correlation between sketch quality and modelling performance in the Pythagorean theorem and a low correlation in Linear Functions; (3) for both topics there is a positive correlation between the representation of the mathematical model in the sketch and modelling performance; (4) the prompt “Make a drawing” (scaffolded by strategy-related hints) does not improve modelling performance neither for Linear Functions nor for Pythagorean theorem when working on modelling tasks. The lack of positive effects of the sketch-prompt is unexpected due to the positive relationship between quality of drawings and modelling performance. We addressed this and other findings in the discussion.