Due to its small size, low weight, and low power consumption, the Rb atomic frequency standard (RAFS) is routinely the first choice for atomic timekeeping in space. Consequently, though the device has very good frequency stability (rivaling passive hydrogen masers), there is interest in uncovering the fundamental processes limiting its long-term performance, with the goal of improving the device for future space systems and missions. The ac Stark shift (i.e., light shift) is one of the more likely processes limiting the RAFS' long-term timekeeping ability, yet its manifestation in the RAFS remains poorly understood. In part, this comes from the fact that light-shift induced frequency fluctuations must be quantified in terms of the RAFS' light-shift coefficient and the output variations in the RAFS' rf-discharge lamp, which is a nonlinear inductively-couple plasma (ICP). Here, we analyze the light-shift effect for a family of 10 on-orbit Block-IIR GPS RAFS, examining decade-long records of their on-orbit frequency and rf-discharge lamp fluctuations. We find that the ICP's light intensity variations can take several forms: deterministic aging, jumps, ramps, and non-stationary noise, each of which affects the RAFS' frequency via the light shift. Correlating these light intensity changes with RAFS frequency changes, we estimate the light-shift coefficient, κLS, for the family of RAFS: κLS = −(1.9 ± 0.3) × 10−12/%. The 16% family-wide variation in κLS indicates that while each RAFS may have its own individual κLS, the variance of κLS among similarly designed RAFS can be relatively small. Combining κLS with our estimate of the ICP light intensity's non-stationary noise, we find evidence that random-walk frequency noise in high-quality space-borne RAFS is strongly influenced by the RAFS' rf-discharge lamp via the light shift effect.