Ultra-processed food: what the label actually covers
The NOVA classification is not about how healthy a food is. It is about how it was made, and that distinction is why hummus and a candy bar can land in different groups than you would guess.
"Ultra-processed" is now in the US dietary guidelines, on packaging, and in a great deal of advice that does not define it. It has a specific meaning, and it is not the one most people assume.
It describes processing, not healthiness
The NOVA system sorts food by how it was made: in four groups:
- Group 1, unprocessed or minimally processed. Fruit, vegetables, meat, eggs, milk, dried beans, plain oats. Cutting, freezing, drying and pasteurizing do not move a food out of this group.
- Group 2, culinary ingredients. Oil, butter, sugar, salt, flour. Things you cook with rather than eat alone.
- Group 3, processed foods. Group 1 plus group 2: canned beans in brine, cheese, bread made from flour, water, salt and yeast, tinned fish in oil.
- Group 4, ultra-processed. Industrial formulations containing little intact whole food, plus substances not used in home cooking: protein isolates, modified starches, hydrogenated oils, emulsifiers, artificial colors, flavorings, non-sugar sweeteners.
The practical test is the ingredient list. If it contains several things you could not buy in a supermarket as themselves, it is probably group 4.
The ones that surprise people
Sitting in group 4: most packaged breads, breakfast cereals, flavored yogurts, plant-based meat substitutes, protein bars, instant noodles, most "diet" products, and nearly everything sweetened with a non-sugar sweetener.
Sitting outside it: butter, cream, sugar, lard, tinned sardines, plain popcorn, and a home-baked cake made from flour, eggs, butter and sugar.
That last one is the sticking point, and it is a fair criticism. A homemade cake is not health food, and NOVA does not pretend otherwise, because NOVA is not ranking healthiness. It is describing industrial processing, on the argument that processing does something to food beyond changing its nutrients.
What the 2026 evidence shows
An umbrella review published in Nutrition Reviews this year synthesized the meta-analyses across obesity, type 2 diabetes, cardiovascular disease, fatty liver, cognition, depression, cancer and all-cause mortality. Higher ultra-processed intake was associated with elevated risk across essentially every outcome examined, with a dose-response pattern, more, worse.
The stronger evidence is the small set of controlled feeding trials. Given free access to matched diets, people eat several hundred extra calories a day on the ultra-processed one and gain weight, without reporting greater hunger. That is a mechanism, not just a correlation.
Where the criticism is legitimate
Two real objections, worth holding alongside the findings:
The category is enormous. It covers a soft drink and a whole-grain supermarket loaf. If those carry different risk, lumping them together obscures more than it reveals.
The observational data cannot separate processing from everything else. People who eat a lot of group 4 also tend to eat less fiber, less produce, more calories and more sodium, and to differ in income, activity and smoking. Studies adjust for what they measure; nobody adjusts perfectly.
The finding is robust. The mechanism, whether it is the processing itself, or what processed food displaces, is still argued about.
What to do with it
Not "eliminate ultra-processed food". That advice fails in week two and ignores cost, time and where most people actually shop.
Something more useful: look at what the ingredient list tells you the food is for. Emulsifiers and modified starch are there to make something shelf-stable and hyper-palatable. That is not a moral failing, but it is a reason a serving is easier to eat quickly and harder to stop eating.
If you track, the thing to watch is not a processing score; it is the things processing predicts. Added sugar, sodium and fiber move together with group 4 membership, and unlike "ultra-processed", they are numbers you can put a cap on.
Common questions
- What is an ultra-processed food?
- Under the NOVA system it is a manufactured product containing little whole food and one or more ingredients you would not find in a home kitchen, modified starches, protein isolates, emulsifiers, high-fructose corn syrup, artificial colors or flavors. The test is the ingredient list, not the calorie count.
- Is all processed food bad for you?
- No. NOVA has four groups, and only group 4 is 'ultra-processed'. Canned beans, frozen vegetables, pasteurized milk, plain yogurt and whole-grain bread are processed and sit in groups 2 or 3. The evidence associating harm with processing is about group 4 specifically.
- Does ultra-processed food cause weight gain?
- Controlled feeding studies show people eat several hundred more calories a day on an ultra-processed diet than on a matched unprocessed one, without reporting more hunger. Large cohort studies link higher intake to obesity, type 2 diabetes, cardiovascular disease and higher all-cause mortality, though those are associations rather than proof of cause.
Sources
Keep reading
The new US dietary guidelines: what actually changed
The 2025–2030 guidelines raised the protein recommendation by roughly half, replaced the added-sugar percentage with a per-meal gram limit, and kept the saturated fat cap exactly where it was, while flipping the food pyramid to feature the foods that cap restricts.
Added sugar: the 10 grams per meal rule
The guidelines replaced a percentage-of-calories limit with a number you can check while holding the jar. It is a better rule, and it will surprise you where it lands.
How much protein do you actually need?
The 2026 guidelines nearly doubled the official figure to 0.54–0.73 g per pound. Here is where that lands for your body weight, and the four situations where you should aim at the top of the range.