In this paper, an adaptive continuous sliding mode (SM) control approach is proposed for buck converters by introducing a novel zero-crossing checking mechanism into the twisting algorithm. Instead of the traditional first-order SM approaches, the twisting algorithm can solve their inherent chattering problem and realize the control continuity at the price of unnecessary constant control gain and low precision. Differing from the traditional adaptive mechanisms based on fuzzy logic or Lyapunov stability, the convergence characteristics of the controlled buck converter system are analyzed and further an adaptive twisting algorithm is proposed based on zero-crossing checking online to achieve a time-varying control gain. The number of the zero-crossing checking points can be calculated and the system stability is investigated. Comparative simulations with the traditional twisting algorithm validate the improved algorithm with advantages of high accuracy and excellent performance.