Non-symbolic numerical stimuli play a crucial role in numerical cognition. Physical properties such as surface area, density, and item circumference are inherently correlated with quantity. The correlations between physical properties and quantity mask the mechanism underlying numerical perception. Non-symbolic stimuli are generated using different generation methods (GMs) aimed at controlling physical properties. The way a GM controls physical properties affects numerical judgments. Here, using a novel data-driven approach, we provide a methodological review of non-symbolic stimuli GMs developed since 2000. Annotators tagged the GMs’ control over physical properties. Next, GMs were qualitatively analyzed for different property controls, terminology, and definitions. The tagging and qualitative analysis provided data suitable for quantitative analysis of GMs. We found that the field thrives with a wide variety of GMs aimed at tackling new methodological and theoretical ideas. However, the field lacks a common language and a method to incorporate new ideas in the existing literature. Furthermore, these shortcomings impair the comparison, replication, and reanalysis of previous studies in light of new ideas. We present guidelines for GMs that will hopefully contribute to the field. First, researchers should define controlled properties explicitly and consistently. Second, researchers should provide the code package used to generate stimuli. Third, researchers should also provide the actual stimuli, coupled with the behavioral and neuronal responses. This last guideline would enable researchers to reanalyze previously obtained data, enabling incorporating new ideas in the context of prior research.