Food swap recommendations improve healthfulness and reduce carbon footprint of online grocery purchases
A randomized trial of 1,201 US adults found that algorithmically-suggested swaps—recommending substitutes for single food items—modestly improved the nutritional quality of online grocery selections and, depending on the type of swap, reduced their carbon footprint.
Researchers tested whether simple food-swap recommendations could nudge consumers toward healthier and lower-carbon grocery choices in a naturalistic online shopping environment. Between May and June 2025, 1,201 US adults (ages 18+) were enrolled in a randomized trial with four arms: climate-focused swaps, health-focused swaps, combined climate-and-health swaps, and a control group that received no recommendations.
Participants completed three weekly shopping tasks; the first established a baseline without any intervention, and the second and third incorporated their assigned recommendations. The study measured two primary outcomes using objective metrics: the Ofcom Nutrient Profiling Model score (1–100 scale, higher scores indicating less healthy choices) and the carbon footprint (kg CO2-equivalents per kg of food). Health-focused swaps reduced the healthfulness score by 4.7 points compared to control (95% CI [3.3, 6.0], p < 0.001), while climate-focused swaps achieved a 2.4-point improvement (95% CI [1.1, 3.7], p < 0.001). Climate-focused and combined swaps both reduced carbon footprint by approximately 3 kg CO2-eq per kg (p < 0.001), whereas health-focused swaps alone did not significantly reduce carbon impact (p = 0.51). The effects did not vary meaningfully by age or environmental consciousness. A key limitation was that the sample skewed somewhat from the broader US population, and results from an online store may not translate to traditional brick-and-mortar shopping.
The study's secondary findings reveal important trade-offs: health-focused messaging alone did little to reduce carbon footprint, suggesting that environmental and nutritional goals may require separate or jointly-framed strategies to achieve both simultaneously. Combined climate-and-health swaps achieved similar carbon reductions to climate-only swaps (B = -2.8 vs. -3.3 kg CO2-eq per kg; p = 0.43) while also improving healthfulness nearly as much as health-only swaps (B = 4.0 vs. 4.7; p = 0.34), making them a promising middle-ground approach. The effect sizes were consistent across demographic subgroups (age, health consciousness, environmental consciousness all p ≥ 0.10), suggesting broad applicability. However, the mechanism likely depends on whether recommended swaps exist in the store's inventory that satisfy both criteria—the study did not report how many swaps were offered, accepted, or what types of foods were actually recommended. The unmasked design (participants knew which arm they were in) may have introduced some behavioral bias, and the online-only setting, where curation and search differ from physical stores, may limit real-world relevance to how most people still shop for groceries.
The deeper explanation, the effect sizes and caveats, and what this means for your own sleep, exercise and reading — read it inside Cadence.
Open in Cadence →