We aim to determine the influence of the Milky Way's warp on the kinematics of stars across the disc, and therefore measure its precession rate and line of nodes under different assumptions. We applied Jeans' first equation to a model of a rigidly precessing warp. The predictions of these models were fitted to the average vertical velocities of stars with measured line-of-sight velocities in DR3 data. We tested models in which the warp's line of nodes and precession speed are fixed, and models in which they are allowed to vary linearly with radius. We also tested models in which the velocity of stars radially in the disc is included in Jeans' equation. The kinematic data are best fit by models with a line of nodes that is $40^ offset from the Sun's Galactic azimuth, significantly leading the line of nodes found from the positions of stars. These models have a warp precession speed of around $13 $ in the direction of Galactic rotation, close to other recent estimates. We find that including the velocity of stars radially in the disc in our kinematic model leads to a significantly worse fit to the data, and implausible warp parameters. The Milky Way's warp appears to be rapidly precessing, but the structure and kinematics of the warped disc are not consistent within the approximation of a fixed, precessing, warp shape. This implies that the Milky Way's warp is dynamically evolving, which is a challenge to models of the warp's creation, and must be considered in the context of other known disturbances of the disc.