Pick up two loaves of bread. One lists four things: flour, water, yeast, salt. The other lists twenty-two, and somewhere past the halfway mark you hit words like mono- and diglycerides, datem, calcium propionate, and soy lecithin. Both are sold as "bread." But they are not the same kind of food, and the difference isn't the calorie count on the back panel. It's what had to happen in a factory to make them.
That distinction has a name. The NOVA system, developed at the University of São Paulo and now used by the WHO and UN, sorts food by how much industrial processing it has been through rather than by its nutrients. The top rung is Group 4: ultra-processed food, or UPF. The tell for Group 4 isn't a single bad nutrient. It's the presence of marker ingredients: additives that exist only because industry put them there.
This matters because ultra-processed food is now the majority of calories in many diets, and it behaves differently in the body than the whole foods it's built from. Nutrication takes it seriously enough to penalize it hard: on the processing measure, a Group 4 food earns just 4 out of 35 points, versus a full 35 for a minimally processed one. That's roughly an 89% haircut for being ultra-processed, before any other measure weighs in.
What actually counts as "ultra-processed"
NOVA (attributed in our engine to Monteiro et al., 2019) uses four groups. The line that trips people up is between Groups 3 and 4, so it helps to see all four together with concrete examples.
| Group | What it is | Examples | Nutrication points (of 35) |
|---|---|---|---|
| 1 | Unprocessed / minimally processed | Apple, plain oats, eggs, dried lentils, fresh fish | 35 |
| 2 | Processed culinary ingredients | Olive oil, butter, salt, sugar, vinegar | 28 |
| 3 | Processed foods | Canned beans, fresh cheese, tinned sardines, real bread | 18 |
| 4 | Ultra-processed (UPF) | Snack cakes, instant noodles, soda, chicken nuggets, most breakfast cereal | 4 |
When a food database hasn't already supplied a NOVA group, the engine draws the line by scanning the ingredient list. A food with two or more recognized industrial additive markers is graded Group 4; exactly one marker makes it Group 3; a food with only simple "processed" ingredients such as salt, sugar, oil, butter, vinegar, yeast, or baking soda lands at Group 2; and a food with none of these is Group 1. A long ingredient list is itself a signal: when a Group 4 food carries no catalogued additives but more than eight ingredients, the engine explains the rating by the length alone — "characteristic of an industrially formulated product." For a deeper tour of the four tiers, see the NOVA processing levels.
The marker-ingredient idea
Here's the useful mental model. Most additives are not poisons. Our additive reference is blunt about this: the stance across its entries is that a given additive is usually "benign but a telltale marker of an ultra-processed food" — the signal matters more than the toxicity. You don't need to memorize the safety file on each one. You need to notice that they're there at all, because their presence means the food was engineered, not cooked.
The markers cluster into a few recognizable families:
- Emulsifiers that keep oil and water from separating: soy lecithin, mono- and diglycerides, polysorbate 80.
- Thickeners and stabilizers for texture: carrageenan, cellulose gum (carboxymethylcellulose), modified starch.
- Cosmetic colors and flavors: tartrazine (Yellow 5), Allura Red (Red 40), "artificial flavor," "natural flavor."
- Industrial sweeteners and fillers: high-fructose corn syrup, maltodextrin (whose glycemic index runs higher than table sugar), protein isolates.
When Nutrication grades a Group 4 food, it names the specific markers it found (up to four, plus a count of the rest) and calls them "cosmetic and shelf-life additives… the hallmark of ultra-processing." Some of these overlap with genuine red flags worth avoiding, which we cover in ingredient red flags. But most are simply evidence, not danger.
Why ultra-processed food is so easy to overeat
The strongest reason to care about UPF isn't a scary chemical. It's appetite. In a tightly controlled feeding study run by the US National Institutes of Health, people given an ultra-processed diet ate roughly 500 more calories a day than the same people eating a minimally processed diet matched for sugar, fat, fiber, and salt — and they gained weight on it. The food itself drove the overeating.
Why? Several ordinary properties stack up. Ultra-processed food is usually energy-dense — lots of calories in a small, dry, low-water package. It's soft and easy to chew, so people eat it faster and register fullness later. It's engineered to be hyperpalatable, hitting fat, sugar, and salt combinations that rarely occur in nature. And it's often low in the fiber and intact structure that make whole foods filling. Drinks are the extreme case, because liquid calories bypass fullness signals almost entirely — see liquid calories. None of this requires a conspiracy. It's just what happens when food is optimized to be cheap, shelf-stable, and moreish.
How to spot it in 5 seconds
Ignore the front of the package. "Natural," "protein," "made with real fruit," and a photo of a wheat field tell you nothing about processing. Flip it over and read the ingredient list, which is the only honest part of the label. Three quick tells:
- Length. A short list of things you could buy separately is a good sign. A list that runs long — well past eight items — leans ultra-processed.
- Unfamiliar names. If several ingredients are words you'd never keep in a kitchen (emulsifiers, gums, isolates, dyes, "flavor"), you're holding a Group 4 food.
- Order. Ingredients are listed by weight. Sugar, refined oil, or a syrup near the top tells you what the food mostly is.
If you want to get fluent at reading the panel itself, how to read a nutrition label walks through it line by line.
The purity trap
Here's what the research does not say: that you must eliminate every ultra-processed food. Processing is a spectrum, not a sin, and treating it as one is both miserable and counterproductive. Canned beans, whole-grain bread, plain yogurt, and tinned fish are lightly processed and genuinely good for you. Some UPF — a protein bar on a long drive, a frozen meal on a hard week — earns its place.
A realistic strategy targets frequency, not purity. Identify the two or three ultra-processed items you eat most days — the soda, the packaged snack, the sweetened breakfast cereal — and swap those for a less-processed default. Leave the occasional treat alone. Cutting your everyday UPF baseline matters far more than agonizing over a birthday cake, and it's a change you can actually keep.
How Nutrication uses this
Processing is the single heaviest input to a Nutrication grade: NOVA contributes up to 35 of the 135 raw points, more than any other measure — though measures without data are simply left out and the rest are rescaled to 100, so nothing is ever guessed. Group 4 collects just 4 of those 35, which is enough on its own to put an A grade out of reach: even a UPF that aced every other measure would land in B territory, and in practice the knock-on penalties below usually drag it to C or D. A Group 4 food also triggers a plain warning on the result card: "Ultra-processed food."
But the engine tries hard not to be dumb about it. If a database has tagged something Group 4 while its ingredient list is short (five items or fewer), free of industrial additives, and clearly a recognizable whole food — canned salmon in oil and salt, say — the grade is quietly reconciled down to Group 3, and the app tells you so: "Some databases tag it 'ultra-processed,' but a recognizable, short ingredient list is more consistent with a processed food — so we've graded it that way." Snacks and chips are deliberately not given that mercy.
Being ultra-processed also shadows the other measures. In nutrient density, a Group 4 food's micronutrient credit is halved, on the logic that "a vitamin-fortified sugary cereal shouldn't read as nutrient-dense as spinach." On fat quality, a fried Group 4 snack has its fat score cut in half too — "a bag of chips shouldn't earn the same fat credit as a handful of almonds just because the fryer used sunflower oil." And when a food is data-poor, being processed lowers the ceiling: a grade resting on only two measures is capped at C if the food is Group 3 or 4. One principle holds throughout, described in how we grade: if there's no ingredient list at all, the processing measure abstains rather than guess. The World Health Organization uses NOVA in its own guidance, which is part of why we anchor to it.
What to actually do
- Read the back, not the front. The ingredient list is the fastest UPF detector you own.
- Count and recognize. Long list plus unfamiliar names equals ultra-processed. That's the whole 5-second test.
- Fix your everyday defaults first. Swap the two or three UPF items you eat most days; the rest matters far less.
- Don't fear light processing. Canned beans, plain yogurt, tinned fish, and real bread are friends, not enemies.
- Watch drinks hardest. Ultra-processed liquids overshoot appetite more than any solid.
- Aim for a better baseline, not perfection. Purity isn't the goal; frequency is.