In this paper, a weak frequency jump detection method is developed for onboard clocks in global navigation satellite systems (GNSS). A Kalman filter is employed to facilitate the onboard real-time processing of atomic clock measurements, whose N-step prediction residuals are used to construct the weak frequency jump detector. Numerical simulations show that the method can successfully detect weak frequency jumps. The detection method proposed in this paper is helpful for autonomous integrity monitoring of GNSS satellite clocks, and can also be applied to other frequency anomalies with an appropriately modified detector.