<p>We present a method for improving next-to-leading order electroweak (EW) predictions for lepton-scattering processes by consistently including double neutral-current corrections arising from vector-boson-fusion topologies, which are formally of higher order. By combining, in a process-independent manner, exact fixed-order results, collinear resummation of QED radiation, and a subtraction procedure, we obtain results which are gauge invariant and valid in the entire phase space, retain any dependence on the masses of electroweak bosons, and can be systematically improved, while avoiding the need for complete next-to-next-to-leading order calculations. This paper is devoted to the development and validation of the formalism; phenomenological applications are presented in a companion study, where we also discuss and motivate why our approach is superior to the one based on EW parton distribution functions for targeting percent-level precision at multi-TeV lepton colliders.</p>