The ability to map ocean-floor magnetization is key to infer past plate motions. The advent of geodesy in the Earth Sciences offered an independent snapshot of contemporary plate kinematics, averaged over decades. Early studies suggested plate-motions steadiness through geological time, because contemporary rates were similar to past 3-Myr averages. Recent data, however, show that geodetic and paleo-magnetic motions exhibit differences beyond confidence ranges, which might suggest geodesy samples time scales shorter than a few centuries. These differences pose important questions on the character of plate kinematic variability through time. Here I derive inferences on the steadiness of geodetically derived plate motions by combining geodynamical arguments with kinematic data sets. I exploit data sets to build scenarios for the recent evolution of most of the major plates. I compare the minimum rate at which torque needs to vary to generate these scenarios, with the maximum rate at which geological processes can contribute torque. This allows assessing the geodynamical plausibility of each scenario. Results indicate that plate-motion changes recorded since ∼3 Ma must have taken at least 1 Myr to occur. Two exceptions are the Pacific and Australian plates, whose motions changed by less than 5%, and therefore, required torque variations that might be built over periods as short as 0.1 Myr. It remains unresolved for how long geodetically derived plate motions kept steady in the geological past. However, results indicate it is indeed plausible that they did over the past 1–2 Myr. These inferences call for a shift in the way we regard the figure of geodetic plate motions.