News · Nutrition & Diet
Climate labels on menus shifted what 10,000 people ordered
A JAMA Network Open randomized trial gave 10,342 US adults a fast-food menu with one of six label designs and found graded climate labels cut selection of high-impact items by up to 8 percentage points.
Based on a peer-reviewed randomized trial in JAMA Network Open
- Researchers at Johns Hopkins and collaborators ran an online randomized trial with 10,342 US adults, publishing in JAMA Network Open.
- Participants ordered from a menu modelled on a US fast food chain under one of six label conditions, including a climate grade from A to F and a red warning label.
- In the control group, 53% chose a high climate impact item. Every graded or warning label beat that: 48%, 46%, 45% and 47% respectively.
- Emissions of the average order fell from 18.8 kg of carbon dioxide equivalent in the control to 16.3 with the grade-plus-scale label.
- Overall nutrition scores did not shift, but the grade-plus-scale group chose items with less saturated fat (11.8 g vs 12.6 g).
- A label showing environmental cost in dollars changed nothing measurable, despite people rating it the most off-putting when shown on its own.
- This was hypothetical online ordering, not a real restaurant. What people click is not always what they buy.
Calorie counts on menus have been law in the United States since 2018, and their effect on what people order has been famously underwhelming. A trial in JAMA Network Open, published by the American Medical Association, tested a different number entirely.
Researchers showed 10,342 US adults a fast-food menu carrying one of six label designs and watched what they ordered. Labels that graded a meal’s climate impact moved choices. A label that priced it did not.
Why a climate label is also a health question
The premise linking the two is stated up front: food systems account for approximately one-third of global greenhouse gas emissions, and many foods with high emissions pose risks to human health. Beef sits at the top of both lists.
Until now that overlap had been argued more than tested, which is what makes this more than an environmental study. If a climate label steers people away from the same items a cardiologist would, it does two jobs at once. Whether it actually does was the open question.
Six labels, one menu
The design was an online randomized clinical trial using a menu based on a US fast food chain, run between December 2024 and February 2025.
Participants were randomized to 1 of 6 label conditions: a control with no label, an environmental cost shown as money, a climate grade from A to F, that same grade shown alongside the full scale, a plain high climate impact flag, and a red warning version of it.
The study included 10 342 participants, average age 46.4 years, 53% women.
The labels that worked
In the control condition, 920 of 1728 people, or 53%, picked an item with a high climate impact. Participants in the climate grade, climate grade plus scale, high climate impact, and warning high climate impact conditions were less likely to select an item with a high climate impact than those in the control condition, at 48%, 46%, 45% and 47%.
Emissions followed. Orders in the grade plus scale condition averaged 16.3 kg of carbon dioxide equivalent against 18.8 kg in the control, with the high climate impact flag at 16.9.
The nutrition story is thinner. Nutrient Profile Index scores did not differ among label conditions. Only one nutrient moved: participants in the climate grade plus scale condition selected items with less saturated fat, 11.8 g versus 12.6 g.
The label that people liked and ignored
The dollar-cost label produced the study’s most interesting failure. There were no significant differences between the environmental cost and control labels for primary or secondary nutrition outcomes.
Yet when people were shown labels on their own and asked to judge them, they perceived the environmental cost label as the most discouraging of selecting high climate impact items.
People said it would put them off. It did not put them off. That gap between stated preference and behavior is a standing hazard in this literature, and this trial caught it cleanly.
How much weight it carries
The obvious limit is that nobody ate anything. This was an online menu with no money, no hunger and no queue behind you, and hypothetical ordering reliably overstates how much information changes behavior.
The effects are also modest in absolute terms: a shift of five to eight percentage points in the share choosing a high-impact item, from a baseline where more than half did.
Still, five points across a national chain is a large number of meals, and the design was properly randomized with a sample big enough to separate six conditions. The authors’ conclusion stays measured, suggesting climate labels, particularly climate grade plus scale designs, may decrease climate impact and improve nutrition of restaurant orders, while calling for tests in real restaurants.
The lesson for anyone designing one: a grade works better than a price, and a grade you can compare works better than a grade alone.
People also ask
Did the labels actually make people eat better?
Barely. The primary effect was environmental: fewer high-emission items, lower emissions per order. On nutrition, Nutrient Profile Index scores did not differ among label conditions. The one nutritional signal was that the grade-plus-scale group chose items with less saturated fat, 11.8 g against 12.6 g. So this is mainly a climate intervention that happens to nudge one nutrient.
Which label design worked best?
The climate grade plus scale, meaning a letter grade shown alongside the full A-to-F range so people could see where an item sat. That condition produced both the lowest share of high-impact selections and the lowest emissions per order. Context appears to matter: a grade on its own is less informative than a grade you can place against alternatives.
Why did the dollar-cost label fail?
That is the study's oddest result. Participants rated the environmental cost label as the most discouraging when viewing labels in isolation, yet it produced no significant difference from control on the actual ordering outcomes. Stated preference and behavior came apart. The authors call for more work on that design rather than writing it off.
Does an online experiment tell us anything about real restaurants?
Less than a real restaurant would, and that is the main limitation. People ordered from a simulated menu with no money at stake, no queue, no hunger and no social context. Hypothetical choices generally overstate how much labels change behavior. The authors explicitly flag the need for climate label use in real restaurant settings.
What counts as a high climate impact item?
Mostly beef and lamb, which carry far higher greenhouse gas emissions per serving than chicken, fish or plant-based options. Food systems account for roughly one-third of global greenhouse gas emissions, and the researchers note that many foods with high emissions also pose risks to human health, which is why the two agendas overlap.