This paper reviews 121 studies on tourism demand modelling and forecasting published since 2000. The latest developments in quantitative forecasting techniques are categorized into three areas: time series models, the econometric approach, and other emerging methods such as AI techniques. Although recent studies show that newer and more advanced forecasting techniques can improve forecast accuracy under certain conditions, no clear evidence shows that any one model consistently outperforms others in forecasting competitions. The review highlights that tourism demand forecasting methods have become more diverse than previously identified. In addition to traditional time series and econometric models, new techniques such as artificial intelligence have emerged. However, no single model consistently outperforms others in all situations. The study identifies new research directions, including improving forecast accuracy through forecast combination, integrating qualitative and quantitative approaches, tourism cycle and seasonality analysis, crisis impact assessment, and risk forecasting. Tourism demand modelling and forecasting rely heavily on secondary data. The most popular measure of tourism demand is tourist arrivals, which can be decomposed into different categories. Other variables such as tourist expenditure, tourism revenues, and employment are also used. The availability of data determines the geographical coverage of forecasting models. The USA, UK, and France are the most researched countries, while Asia has gained increasing interest due to its fast growth. Quantitative forecasting techniques, including time series models and econometric approaches, are widely used. Time series models such as ARIMA and SARIMA are dominant, but their performance varies. Econometric models, including VAR, ECM, and ADLM, are also used. These models often outperform time series models in forecasting accuracy. However, the performance of these models can vary depending on the data frequency, destination, and forecasting horizon. Artificial intelligence techniques, such as neural networks and support vector machines, have also been applied in tourism forecasting. These methods often outperform traditional models in certain cases. However, they lack theoretical foundations and are difficult to interpret. Forecast combination and integration of quantitative and qualitative approaches have been suggested as ways to improve forecast accuracy. However, the effectiveness of these methods is still under investigation. The study also highlights the importance of seasonality analysis, crisis impact assessment, and risk forecasting in tourism demand modelling. In conclusion, while there is no single model that consistently outperforms others in all situations, recent studies show that new techniques and methods can improve forecast accuracy. Further research is needed to explore the effectiveness of forecast combination, integration of qualitative and quantitative approaches, and the handling of seasonality in tourism demand forecasting.This paper reviews 121 studies on tourism demand modelling and forecasting published since 2000. The latest developments in quantitative forecasting techniques are categorized into three areas: time series models, the econometric approach, and other emerging methods such as AI techniques. Although recent studies show that newer and more advanced forecasting techniques can improve forecast accuracy under certain conditions, no clear evidence shows that any one model consistently outperforms others in forecasting competitions. The review highlights that tourism demand forecasting methods have become more diverse than previously identified. In addition to traditional time series and econometric models, new techniques such as artificial intelligence have emerged. However, no single model consistently outperforms others in all situations. The study identifies new research directions, including improving forecast accuracy through forecast combination, integrating qualitative and quantitative approaches, tourism cycle and seasonality analysis, crisis impact assessment, and risk forecasting. Tourism demand modelling and forecasting rely heavily on secondary data. The most popular measure of tourism demand is tourist arrivals, which can be decomposed into different categories. Other variables such as tourist expenditure, tourism revenues, and employment are also used. The availability of data determines the geographical coverage of forecasting models. The USA, UK, and France are the most researched countries, while Asia has gained increasing interest due to its fast growth. Quantitative forecasting techniques, including time series models and econometric approaches, are widely used. Time series models such as ARIMA and SARIMA are dominant, but their performance varies. Econometric models, including VAR, ECM, and ADLM, are also used. These models often outperform time series models in forecasting accuracy. However, the performance of these models can vary depending on the data frequency, destination, and forecasting horizon. Artificial intelligence techniques, such as neural networks and support vector machines, have also been applied in tourism forecasting. These methods often outperform traditional models in certain cases. However, they lack theoretical foundations and are difficult to interpret. Forecast combination and integration of quantitative and qualitative approaches have been suggested as ways to improve forecast accuracy. However, the effectiveness of these methods is still under investigation. The study also highlights the importance of seasonality analysis, crisis impact assessment, and risk forecasting in tourism demand modelling. In conclusion, while there is no single model that consistently outperforms others in all situations, recent studies show that new techniques and methods can improve forecast accuracy. Further research is needed to explore the effectiveness of forecast combination, integration of qualitative and quantitative approaches, and the handling of seasonality in tourism demand forecasting.