At the point of stall most aircraft are likely to drop one wing or the other, either due to the presence of yaw during the approach to the stall, or otherwise because no aircraft is built with perfect symmetry.
If yaw is not present and the angle of attack is allowed to increase, the wing drop is likely to be short lived with the other wing stalling very soon thereafter. At this point the aircraft becomes fully stalled, and the angle of bank will stabilize in the direction of the wing that stalled first (still assuming no yaw is present). In other words, and unless the angle of attack is increased, rotation (one wing stalled while the other generates lift) is not present or a primary concern.
Accordingly, as has been pointed out by others here, reducing the angle of attack of both wings simultaneously with elevator should be one's first concern, and then when the aircraft is un-stalled should one address leveling the wings. In this scenario using opposite aileron at the point of stall to level the wings, depending on the type, may make matters worse. Doing so generally increases the angle of attack on the wing that stalled first and reduces it on the wing that is still flying, and could lead to the development of a rotation. Unloading both wings with a pitch down is therefore a priority, but as those who are familiar with outside flick rolls will tell you the unloading with forward pitch should be done at a relatively gentle pace, and certainly not aggressively.
Off topic, but in jet types and when flying at high subsonic mach numbers, the angle of attack at which the wing stalls is substantially less than the angle at which the wing would stall at lower speeds/levels. Accordingly those flying older types without air data logic that correct for this anomaly, should be careful when operating close to ceiling limitations. Old style AoA indicators can mislead by as much as 30% when passing M0.75, and fool one into believing that they have a bigger stall margin than really exists.