The aim is to highlight the complexity for teachers to conduct screening of spatial reasoning to identify children at risk at an early stage. Spatial reasoning is crucial in developing children’s mathematical knowledge (Clements and Sarama, 2011). Preschool teachers may direct teaching toward the screening activity (Walla, 2024), which is known as teaching for the test (Volante, 2004). This implies that assessment results do not necessarily reflect children’s spatial reasoning outside the screening situation. To understand teacher-child interaction in the screening activity, the lens of legitimation code theory (LCT) is used to examine what role the teacher plays in identifying children atrisk. This qualitative study uses LCT’s semantic profiles as combination of semantic gravity and density to analyse three consecutive 20–26-minute video-recorded screenings conducted by a female teacher with three groups of foursix-year-olds each. The teacher’s, children’s and their guardians’ informed consent was obtained with the right to withdraw without prejudice. Children’s consent was monitored during data collection by observing their non-verbal expressions. The findings suggest that in 83 % of interactions, the teacher maintained the same degree of semantic gravity and density as the previous child’s utterance, thus conducting the screening. However, she sometimes deviated from this pattern and started teaching instead, potentially failing to identify at-risk children. The study underscores the challenge for teachers to be screening administrators, as they may struggle to resist the temptation to turn a screening session into a learning opportunity. The question remains who is to do the screening then.