Given the fast-paced world economy, increasing carbon dioxide (CO2) emissions have caused serious concern regarding global warming. Many studies have examined the nexus between CO2 emissions and economic growth and inconsistent results are provided. Limited evidence suggests that different time periods under investigation could be a source of inconsistency, but there lacks of a thorough analysis. This paper systematically focuses on the evolutionary characteristics of the nexus to explore the influence of time periods. The panel data are reorganized by considering different time periods in various countries, and an intelligent data-driven approach, symbolic regression, is applied to discover models for each period without predefined structure and parameters. Three issues (Does the time period matter? What are the trends in different time periods? What are the applicable scopes of the models?) have been addressed in this paper. The results reveal the evolutionary nature of the nexus and it is verified that an intrinsic time period distinction does generate extrinsic inconsistency. The detailed dynamics of models indicate that the optimal models tend to be more complex as the time periods are extended. For countries with different locations and income levels, the applicable scopes of the four superior models (monotonically increasing, inverted U-shaped, inverted N-shaped, and M-shaped) vary in different time periods. The findings highlight that the effect of time periods cannot be ignored and the models should not be applied unconditionally when conducting empirical research. Finally, several policy implications are provided, which are critical for the balance of economic growth and CO2 emissions, particularly in some developing countries.