Economists’comprehensive judgments and confidence levels are increasingly valuable in today’s environment of heightened economic fluctuations and uncertainty. However, existing economist confidence indexes are not sufficiently regarded due to their low frequency, high compilation costs, and lack of timeliness. This study constructs a monthly Chief Economist Confidence Index and its subindexes (CECI) by leveraging online text data from the “Chief Economists Forum” and employing cuttingedge natural language processing technologies, specifically the TextRank+FinBERT method. The paper finds that the CECI trends consistently with the National Bureau of Statistics’Quarterly Economist Confidence Index, but features stronger timeliness and a higher update frequency. Compared to the confidence indexes of other economic entities, the confidence of chief economists serves as a more effective indicator of the business cycle. In terms of macroeconomic forecasting, the CECI series can significantly enhance the outofsample forecasting performance for key macroeconomic variables. This study represents a valuable attempt to construct business cycle indicators and conduct macroeconomic forecasts using artificial intelligence methods. The methodology can be applied to the construction of other highfrequency business cycle indexes.