We evaluated at-column dilution (ACD) as a possible approach to reduce peak distortion caused by the injection of strong solvents in preparative hydrophilic interaction chromatography (HILIC). Three model compounds, namely uridine, cytidine and guanosine were selected and injected either in conventional or ACD conditions, using various proportions of water and acetonitrile in the sample diluent. Plate number and peak capacity were systematically investigated under both isocratic and gradient elution conditions, respectively. In isocratic conditions, ACD was found to provide higher plate numbers with sample diluent containing more than 40% water, but lower plate numbers with injections of less than 40% water, in comparison to a conventional injection system at the preparative scale. These contrasting results were attributed to the fact that i) efficiency was fundamentally reduced in ACD, due to the low make-up flow rate that was used to flush the injection loop, but ii) the trend was reversed for highly aqueous injection thanks to the capacity of ACD to maintain similar efficiency regardless of the injection solvent. On the other hand, peak capacity was comparable between ACD and conventional systems in gradient elution for injections containing low amounts of water. However, ACD became increasingly advantageous as the proportion of water in the injection solvent and/or the injection volume increased. This was confirmed by the separation of a plant root aqueous extract. Altogether, our findings suggest that ACD is beneficial both in isocratic and gradient modes when strongly aqueous solutions are injected, offering an attractive approach to increase loadability, while preserving peak shapes in preparative HILIC.