We investigate the predictive content of AI cable news-based economic policy uncertainty (TVEPU) for travel and leisure stock returns in the USA, European Union, and China. Using the Rolling Windows Wavelet Quantile Granger Causality technique, we find that the predictive content of TVEPU varies across time, frequencies, and quantiles. When the travel and leisure stock market is bearish or bullish, the predictive content of TVEPU is stronger in the long run. However, when the travel and leisure stock market is normal, the predictability is stronger in the short run. These findings are supported by the outcomes of the network connectedness analysis.