We discuss the sensitivity of theoretical predictions of observables used in searches for new physics to parton distributions (PDFs) at large momentum fraction x. Specifically, we consider the neutral-current Drell–Yan production of gauge bosons with invariant masses in the TeV range, for which the forward–backward asymmetry of charged leptons from the decay of the gauge boson in its rest frame is a traditional probe of new physics. We show that the qualitative behaviour of the asymmetry depends strongly on the assumptions made in determining the underlying PDFs. We discuss and compare the large-x behaviour of various different PDF sets, and find that they differ significantly. Consequently, the shape of the asymmetry observed at lower dilepton invariant masses, where all PDF sets are in reasonable agreement because of the presence of experimental constraints, is not necessarily reproduced at large masses where the PDFs are mostly unconstrained by data. It follows that the shape of the asymmetry at high masses may depend on assumptions made in the PDF parametrization, and thus deviations from the traditionally expected behaviour cannot be taken as a reliable indication of new physics. We demonstrate that forward–backward asymmetry measurements could help in constraining PDFs at large x and discuss the accuracy that would be required to disentangle the effects of new physics from uncertainties in the PDFs in this region.