The aim of the current study is to compare the effect of virtual and physical manipulatives on preschool children's understanding of geometric shapes. A quasi-experimental research design was employed where children in the experimental group used tablet computers to access Augmented Reality (AR)-based virtual manipulatives, while children in the control group used physical manipulatives. The instruction lasted for a period of four weeks at a public primary school in Turkey with 72 children aged five and six years old. Comparison of their understanding of geometric shapes was based on their scores from pre-test and post-test measures of the Geometric Shape Recognition Task instrument. Analysis of the collected data revealed no statistically significant difference between the groups in the circle recognition task, while statistically significant differences were found between the groups in the recognition tasks for triangle, rectangle, and square shapes in favour of the experimental group. In addition to this, although there was an increase in the total scores of both groups, the results showed a statistically significant difference in test scores in favour of the experimental group. The current study also showed that the children had difficulty in categorising geometric shapes when their attributes were changed, therefore providing different forms of geometric shapes to children is very important for improving their conceptual understanding.