The rising prevalence of non-communicable diseases has brought attention to the importance of consuming a healthy diet. One strategy to improve diet quality is through front-of-pack (FOP) nutrition labels. Taking advantage of an online grocery store, we allowed consumers to choose the FOP labels they preferred, and combined this information with real-time feedback on the overall nutritional quality of the shopping basket. We hypothesized that these dynamic food labels with real-time feedback (DFLF) would improve nutritional quality of food purchases. This trial followed a two-arm (no-label control and DFLF) crossover design with 125 participants exposed to each condition once in random order via an online grocery store. A first difference regression model allowed for estimating the unbiased effect of the DFLF on diet quality, measured by the weighted average Nutri-Score (ranging 1 to 5) per serving (primary) and changes in select nutrients and calories. The mean weighted Nutri-Score was 0.4 (12.6%) higher in the DFLF arm (CI: [0.2, 0.6]) relative to the control. The DFLF also decreased the amount of sugar per serving by 0.9 g (CI: [−1.7, −0.0]) and total sugar per shop by 169.5 g (CI: [−284.5, −54.5]). The DFLF features significantly improved nutrition quality relative to no labelling, as measured by average Nutri-Score values. These results shed light on the considerable potential of the online shopping environment to improve diet quality through customization and real time feedback.