Ultra-Processed Foods: What They Are, and What to Eat Instead
Written By Kevin Wallace.
Canning is not the problem. The industrial formulation is.
1. Why this is the first answer
Not all processing is a problem. Canning tomatoes, drying apple slices, and freezing berries is how people have kept a harvest. The foods worth pushing down the plate are the ultra-processed ones: industrial formulations of refined fractions and additives you would not keep in a farmhouse kitchen. They are now more than half of American calories.
When someone at the nursery asks where to start, this is usually the first answer. The classification, the trials, the limits of the trials, and a way to read a label.
2. Four levels, not one smear
The system most researchers use is NOVA, from Carlos Monteiro and colleagues in Brazil. It sorts foods by the extent and purpose of processing, not by a vitamin score. (1)
Drying apples or freezing berries is Group 1. Canning tomatoes with a little salt, or putting up jam, is Group 3. None of that is ultra-processed. The purpose of Group 4, in Monteiro’s account, is a convenient, hyper-palatable product designed to displace other food. (1)
| NOVA group | What it is | Examples |
|---|---|---|
| 1. Unprocessed or minimally processed | Whole foods, cleaned, cut, dried, frozen, fermented, or pasteurized, with nothing added | Fruit and vegetables, dried beans, whole grains, plain oats, nuts, dried fruit with nothing added |
| 2. Culinary ingredients | Pressed, milled, or refined from Group 1, used to cook | Olive oil, salt, sugar, vinegar, maple syrup, flour |
| 3. Processed | Group 1 foods with Group 2 ingredients, to preserve or flavor | Home-canned tomatoes, pickles, salted nuts, bread from flour, water, yeast, and salt |
| 4. Ultra-processed | Industrial formulations, including substances rarely used at home, built to be convenient and hard to stop eating | Soda, most packaged sweets, many breakfast cereals, instant noodles, reconstituted meats, many meat substitutes |
3. How much of it we eat
In August 2021 to August 2023, Americans aged 1 and older got 55 percent of their calories from ultra-processed foods. Youth aged 1 to 18 got 61.9 percent. Adults got 53.0 percent. Among adults that share was 55.8 percent in 2013–2014, so the drift is down, not up, and it is small. The largest sources were sandwiches, including burgers, sweet baked goods, salty snacks, and sugary drinks. (2)
A sandwich lands in the ultra-processed column mostly because of the bread and the processed meat, not because someone sliced a tomato. That is a quirk of the system, and it is worth knowing before the category is treated as one thing.
4. What the trials show
The cleanest human experiment is still Kevin Hall’s, at the NIH Clinical Center. Twenty adults lived in the unit for a month. Two weeks ultra-processed, two weeks minimally processed, in random order. The diets offered were matched for calories, sugar, fat, fiber, sodium, and macronutrients. People could eat as much as they wanted. On the ultra-processed diet they ate 508 more calories a day and gained 0.9 kg, about 2 pounds. On the minimally processed diet they lost about the same. They did not rate the ultra-processed food as more pleasant. They ate it faster, about 48 calories a minute against 31. (3)
That is cause, for overeating, in this setting. It is not a disease trial. One fair critique is energy density. The diets were called similar only after low-calorie beverages were counted, and those drinks had been used in part to deliver fiber. Solid-food density may not have been equal. Eating rate and texture remain the most concrete differences the trial measured.
A second trial, published in 2025, asked a harder question. Fifty-five adults in England with a body mass index of 25 to 40, already getting at least half their calories from ultra-processed food, ate two 8-week diets that both followed the UK Eatwell Guide. One was minimally processed. One was ultra-processed. Both groups lost weight. The minimally processed diet lost more, about one extra percentage point of body weight. Self-reported intake was about 327 calories a day lower on that diet. (4) Meeting the guidelines helped either way. Processing still moved the scale.
5. What the long studies show, and how sure they are
A 2024 umbrella review pooled 45 meta-analyses and almost 10 million people. Higher ultra-processed intake was tied to 32 outcomes. The authors called the evidence “convincing” for about a 50 percent higher risk of cardiovascular death, a 12 percent higher risk of type 2 diabetes per step up the dose, and roughly 50 percent higher odds of anxiety and common mental disorders. Read the grade next to the label. Cardiovascular death was graded very low quality. Diabetes, on the dose-response analysis, was moderate. Anxiety was low. All-cause death, about 21 percent higher, was “highly suggestive” and still low quality. (5) Association, pooled, not proof.
Thirty-four years of U.S. nurses and health professionals, more than 114,000 people, found a smaller mortality difference: 4 percent higher all-cause death in the top quarter versus the bottom, hazard ratio 1.04. No clear link with cancer death or cardiovascular death in that study. The absolute rates were 1,472 and 1,536 deaths per 100,000 person-years. Ready-to-eat meat, poultry, and seafood products carried the association. Sugary and artificially sweetened drinks, dairy desserts, and highly processed breakfast foods followed. (6) The category is not one risk.
6. Why they may do it
- Speed and texture. In the NIH trial the extra 500 calories arrived without extra hunger, and the food went down faster. Soft texture and a dense mix of refined starch, fat, and salt get ahead of fullness.
- The matrix is gone. Fiber and cell walls in an apple or a bowl of lentils slow digestion. Ultra-processed foods are built from starches, sugars, oils, and protein isolates. That structure is the fiber story from the carbohydrate article.
- Some additives change the gut, in a small trial. In healthy people, the emulsifier carboxymethylcellulose lowered bacterial diversity, lowered short-chain fatty acids, and raised abdominal discomfort versus an additive-free diet. One additive, one study. (7)
- They crowd out beans, greens, fruit, and whole grains. A missing food can matter as much as an added one.
7. Where the evidence is weak
Apart from the two feeding trials, this is observational. People who eat a lot of these foods differ in income, smoking, sleep, and exercise. Adjustment is partial. NOVA puts soda, hot dogs, packaged whole-grain bread, and a plant yogurt with a stabilizer in one bin. Fang’s subgroup analysis is the reason not to restrict them equally. Processing is also tangled with sugar, salt, and refined fat. The trials are the part that tries to pull those apart, and they are short.
None of that rescues a soda or a hot dog. It does argue against treating a loaf with five ingredients as if it were a cola.
8. How to spot them
There is still no U.S. label. In August 2026 HHS and USDA sent a proposed federal definition out for final review. The text was not public at this writing. (8) Until a label exists, the ingredient list is the tool.
The kitchen test: would you find this in a home kitchen? Flour, beans, tomatoes, olive oil, salt, and vinegar pass. These do not.
- Modified sugars: high-fructose corn syrup, maltodextrin, dextrose, invert sugar, fruit-juice concentrate used as a sweetener.
- Modified oils: hydrogenated or interesterified oils.
- Protein isolates and hydrolysates, textured vegetable protein.
- Texture agents: carboxymethylcellulose, polysorbate 80, mono- and diglycerides, carrageenan, and a long list of gums.
- Cosmetic additives: artificial colors, flavors, flavor enhancers, sucralose, aspartame.
A long list, a health claim on the front, and a room-temperature shelf life for a food that should spoil are the other tells. Plant-based is not a pass. Many meat substitutes, plant cheeses, and flavored plant yogurts are Group 4. Beans, lentils, whole grains, vegetables, fruit, nuts, and seeds are the foundation. Open Food Facts will assign a NOVA group to many barcodes. It is a convenience, not a laboratory. (1)
9. What to eat instead
Cook a pot of beans once a week. Keep frozen vegetables and rinsed canned beans. Washed fruit where you can see it. Those are still convenience.
| Instead of | Try | Why |
|---|---|---|
| Soda, sweet tea, sports drinks | Water, sparkling water, unsweetened tea | Drinks were among the stronger mortality links |
| Deli meat, hot dogs, sausage | Beans, lentils, hummus | Ready-to-eat meat products carried the mortality association |
| Sweetened cereal | Rolled or steel-cut oats, fruit, nuts | Whole grain, no colors |
| Packaged cookies and cakes | Fruit, or a simple bake with few ingredients | Fewer emulsifiers and refined fats |
| Chips | Nuts, roasted chickpeas, air-popped popcorn | Harder to eat at 48 calories a minute |
| Flavored yogurt | Plain yogurt and fruit | Drops the sugar and the thickeners |
| Frozen dinners | A pot of soup, frozen in jars | Same convenience, known ingredients |
| Mass-market white bread | Flour, water, yeast or starter, salt | Group 4 to Group 3 |
10. What we do here
We process a lot of food. Tomatoes and fruit go into glass mason jars. Apples and vegetables go through the Excalibur dehydrators. The berry harvest goes into the freezer. The Ninja makes a plant-based ice cream whose ingredient list I can say out loud. By NOVA that is Group 1 or Group 3: heat, drying, cold, and sometimes a little salt or sugar. Nothing synthetic on the plants, and nothing ultra-processed in the jar.
I do not think everyone has to eat the way we do. I do think almost everyone would do better with fewer Group 4 foods, and with a look at the additives that come with them.
11. Where we land
- Processing is not the enemy. Ultra-processing is.
- Start with sugary and artificially sweetened drinks, and with processed meats. Those have the strongest human signal.
- Use the kitchen test. If you would not stock the ingredient, the food is probably Group 4.
- Do not trust the front of the box, including the words plant-based.
- Build convenience from beans, frozen produce, and a pot of soup.
- A few swapped servings is progress. The Hall trial was two weeks. A kitchen can be the same length.
12. Sources
(1) Monteiro et al., Public Health Nutrition, 2018. NOVA. <https: data-preserve-html-node="true"//doi.org/10.1017/S1368980017000234>
(2) Williams et al., NCHS Data Brief No. 536, August 2025. <https: data-preserve-html-node="true"//www.cdc.gov/nchs/products/databriefs/db536.htm>
(3) Hall et al., Cell Metabolism, 2019. <https: data-preserve-html-node="true"//www.cell.com/cell-metabolism/fulltext/S1550-4131(19)30248-7>
(4) Dicken et al., Nature Medicine, 2025. <https: data-preserve-html-node="true"//doi.org/10.1038/s41591-025-03842-0>
(5) Lane et al., BMJ, 2024. <https: data-preserve-html-node="true"//www.bmj.com/content/384/bmj-2023-077310>
(6) Fang et al., BMJ, 2024. <https: data-preserve-html-node="true"//www.bmj.com/content/385/bmj-2023-078476>
(7) Chassaing et al., Gastroenterology, 2022. <https: data-preserve-html-node="true"//pubmed.ncbi.nlm.nih.gov/34774538/>
(8) HHS and USDA, proposed ultra-processed-food definition submitted for final review, August 10, 2026. <https: data-preserve-html-node="true"//www.hhs.gov/press-room/hhs-announces-ultra-processed-foods-gras-reforms.html>
