The present study was conducted with the aim of analyzing the role of artificial intelligence (AI) in either reproducing or mitigating educational inequality, and designing a model for "smart educational justice" in schools located in underprivileged areas of Tehran. In terms of purpose, this is an applied study; methodologically, it employs a descriptive-survey design with a correlational and explanatory approach. The statistical population consists of principals, vice-principals, teachers, and educational experts in schools within underprivileged districts of Tehran, with a proposed sample size of 300 individuals. Data were collected using a researcher-developed questionnaire and analyzed via Partial Least Squares Structural Equation Modeling (PLS-SEM). The resulting model indicates a positive relationship between smart educational justice and the following factors: AI-based personalized learning (path coefficient: 0.243), digital infrastructure (0.216), targeted allocation of educational resources (0.194), teacher empowerment (0.187), organizational readiness (0.168), and digital literacy (0.141). Furthermore, the path coefficient from "algorithmic bias" to the reproduction of educational inequality is 0.318, while the path coefficient from "responsible governance" to the reduction of such inequality is 0.276. For this sample size, the coefficient of determination ($R^2$) for smart educational justice is 0.724, and for the reproduction of educational inequality, it is 0.548; this signifies that the model explains 72.4% and 54.8% of the variance in these two constructs, respectively. Furthermore, the interaction coefficients for the application of artificial intelligence with teacher empowerment and digital infrastructure are 0.124 and 0.157, respectively. The Standardized Root Mean Square Residual (SRMR) value in this model is 0.061. In summary, the research’s conceptual framework emphasizes that the potential of artificial intelligence to enhance educational equity depends on equitable access to digital infrastructure, the empowerment of human resources, the personalization of learning, the targeted allocation of resources, and the implementation of responsible governance mechanisms. Confirming these relationships and statistical values ​​requires the collection of field data and the execution of a final analysis.