For in-service GFRP (glass fibre-reinforced polymer) cellular bridge decks, wheel loads can induce damage of the anti-skid surfacing-to-deck bond, thereby loosening fragments of surfacing that then cause abrasion of the deck’s top flange under further wheel loads. Indeed, over the last 20 years, wheel load-induced damage to the top flanges of FRP decks on the road network has been observed in different countries, after periods in service ranging from a few days to a few years, due to inadequate design. Bonding of uni-directional GFRP panels to the abraded multi-directional flange constitutes a potentially effective rehabilitation strategy, but it also influences the deck’s local wheel-load response by altering the anisotropy of the top flange. The present paper uses an experimental-numerical study to explore this role of anisotropy for both abraded and rehabilitated decks. The experiments captured the top flanges’ peak biaxial strains, using gauges installed on the flange soffits to avoid damage by the wheels. The numerical model captures the morphology of the deck, the actual spread of the tyre load over the contact zone and spatial variations in anisotropy of the abraded or built-up flanges and of the webs. The results show that the uni-axially dominated rehabilitation strategy strongly improves bi-axial response to static wheel loads. The numerical model encouragingly predicts the bi-axial strains for each of the abraded and rehabilitated decks, and reliably predicts the strain reductions due to rehabilitation. Refined meshes along the tyre-loaded local flange span, with coarser meshes in adjacent spans, enable these reasonable predictions of response. It is concluded that top flange soffit gauges installed via ad-hoc tooling inserted through holes in the bottom flange, in tandem with rigorous numerical modelling, constitutes an effective two-pronged approach for elucidating wheel load effects.