Aerodynamically, wings generate "lift" normal to the local flow and "drag" parallel to the local flow. (In fact, of course, "lift" and "drag" are just convenient ways to break down the total forces....)
Now, because of the tip vortex the flow which is hitting the winglet isn't parallel to the general airflow - it has a significant spanwise component, as the vortex "rolls inwards over the top of the wing".
That means that the local flow is such that when the winglet acts on it you get a "lift" compoenent from the winglet which is partially inwards, partially upwards and partially
forwards.
Something like this:
DC-M is correct to say its minor, and its very much dependent on the conditions (which is why winglet optimisation is an art, and why winglets are much better on aircraft with a powerful single design point, rather than a range of design points)