<p>This paper introduces a novel online portfolio selection strategy that integrates investors’ attention. Existing research has found that investors’ attention has a significant impact on the stock market. However, traditional strategies typically rely solely on explicit data variables, such as historical relative prices, neglecting crucial implicit variables like investors’ attention and other external factors. To address this, we first use the search index as a proxy variable for investors’ attention to measure the focus level of investors. Second, leveraging the herding effect in behavioral finance, we propose a strategy for transferring investment weights via investors’ attention between pairs of stocks. Finally, based on moving window and expert strategies, we construct an aggregating online portfolio strategy by using the meta algorithm of online gradient update algorithm, which has theoretically guarantees on the sublinear upper bound of regret. Empirically, the strategy outperforms traditional methods in most cases in terms of cumulative wealth and risk-adjusted returns. The annualized return rates for all datasets range from 5% to 31%, with average Sharpe, Calmar and information ratios of 0.4537, 0.5064 and 0.0653, respectively. Furthermore, the strategy demonstrates robustness under different parameters and can withstand reasonable transaction costs.</p>