Graphical abstract Highlights • A rapid technique was achieved for sensing total theaflavins content in black tea. • Total theaflavins prediction models were improved with variable selection algorithms. • Cyclic voltammetry coupled Si-CARS-PLS was most reliable for prediction. Abstract This study attempted to detect the total theaflavins content in black tea using a portable electronic tongue integrated with chemometric algorithms. Glassy carbon as a working electrode was used to collect the cyclic voltammetry current signals from the black tea samples. The synergy interval partial least square with competitive adaptive reweighted sampling (Si-CARS-PLS) was attempted to optimize and select the most informative current signal variables at special potentials for the prediction of the total theaflavins content in black tea. Models were optimized via cross validation. Compared with other characteristic variables selection methods, Si-CARS-PLS showed the best performance, employing only 13 variables (0.23% of the total variables), to achieve R p = 0.8302 and RMSEP = 0.257 in the prediction set. The portable electronic tongue based on glassy carbon electrode and cyclic voltammetry combined with variable selection algorithm Si-CARS-PLS, proved a promising, rapid and cost-effective method to measure the total theaflavins content in black tea. [ABSTRACT FROM AUTHOR]