Artificial Intelligence (AI) models, particularly neural networks, are infrequently utilized in the existing airport management literature for conventional forecasting of airport activities. The limited adoption of these models in the airport management literature might be influenced by their perceived complexity. This perception is likely derived from their common application in intricate tasks within the academic literature. Nevertheless, this research calls for a reevaluation of such perceptions and advocates for the inclusion of RNN and multivariate RNN in the forecasting toolkits of airport managers as credible alternatives to traditional time series models. This study endeavors to discern the forecasting performance of neural network models, providing insights into their effectiveness and applicability in addressing the complexities of passenger flow dynamics through a comprehensive evaluation of RNN, LSTM, GRU, Deep LSTM, BLSTM, multivariate RNN and multivariate LSTM, in comparison to standard time series models (ARIMA, SARIMA and SARIMAX). It was anticipated that the application of neural network techniques in TSA passenger flow v forecasting will yield heightened accuracy when compared to conventional standard time series models. Moreover, the integration of non-standard external factors was expected to enhance the forecasting performance of neural network-based models like RNN and LSTM, further distinguishing them from standard time series models. This investigation rigorously evaluates the robustness of these models by subjecting them to highly volatile historical data to forecast airport security checkpoint passenger flow at five prominent U.S. airports during the pandemic-induced challenges. At Atlanta's Hartsfield-Jackson Airport (ATL), the forecasting precision of RNN notably exceeds that of SARIMA by 34% (DM= 3.44, p< 0.01). This highlights the superior capacity of RNN to manage intricate interactions among variables, complex dependencies between factors and non-linear dynamics, thereby demonstrating its readiness for the emerging data-rich aviation environment. The incorporation of exogenous variables enhances the forecasting accuracies of the multivariate RNN/LSTM (DM=6.82, p