The First US Ultra-Processed Food Definition: What It Means

For a term that appears in headlines almost daily, "ultra-processed food" has never had a legal definition in the United States. Not once. Not in any statute, regulation, or federal guidance document.

That is about to change. On 10 August 2026, the Department of Health and Human Services and the Department of Agriculture jointly submitted the first proposed federal definition of ultra-processed foods to the Office of Management and Budget for final review. As of this writing the text has not been released publicly, and the agencies have not disclosed the criteria or thresholds it contains.

This matters more than a vocabulary debate. A federal definition is the hinge on which a great deal of future policy turns — school meal procurement, front-of-package warning labels, federal nutrition programmes, marketing restrictions, research funding, and the language food companies may legally use on a packet. Once a term has a legal boundary, it can be regulated. Until then, it cannot.

This guide covers what ultra-processed food currently means under the scientific framework everyone has been borrowing, what the American process has looked like so far, what the definition will have to resolve, and — most practically — how to identify ultra-processed food yourself today, without waiting for a regulator to tell you.

Because no legal definition existed, researchers, journalists and policymakers have all been using the same borrowed one: NOVA, developed by a team at the University of São Paulo led by Carlos Monteiro.

NOVA's central insight is a departure from how nutrition has traditionally been assessed. It does not classify food by nutrient content — calories, fat, sugar, sodium. It classifies by the extent and purpose of industrial processing. Two products with near-identical nutrition panels can sit in different NOVA groups, and NOVA's argument is that the difference matters independently of the nutrients.

The operative distinction between Group 3 and Group 4 is not "how much processing" in a vague sense. It is whether the product contains substances you would never find in a domestic kitchen — emulsifiers, stabilisers, humectants, bulking agents, colour retention agents, flavour enhancers, modified starches, protein isolates, hydrogenated oils.

Bread illustrates this cleanly. Bread made from flour, water, salt and yeast is Group 3. Industrial sliced bread containing emulsifiers, preservatives, dough conditioners and added gluten is Group 4. The nutrition panels may look similar. The classification is entirely different.

Writing this definition is genuinely difficult, and the difficulty is not bureaucratic. Four specific problems have no clean answer.

The first is that NOVA was designed for population research, not regulation. It classifies by process and intent, which works well when analysing dietary patterns across a country. It works poorly as a legal test applied to a specific product by a specific manufacturer, because "would a home cook use this?" is not an enforceable standard. A regulation needs criteria a company can apply in advance and an inspector can verify.

The second is that processing and nutritional harm are correlated but not identical. Some Group 4 foods are nutritionally reasonable — a fortified wholegrain cereal with added vitamins, or a plant-based milk with added calcium. Some Group 3 foods are not — salted butter is Group 2, and few would call it health food. A definition that captures fortified cereals while exempting butter will be attacked as incoherent, and the attack will have a point.

The third is the additive threshold question. If the definition turns on the presence of additives "not used in home cooking", it needs a list. Every substance on that list becomes a commercial cliff edge, and every substance near the boundary becomes a lobbying target. If instead it turns on a count of additives, manufacturers reformulate to sit one below the line without changing anything meaningful.

The fourth is scope. A narrow definition catching only the most obviously industrial products will be criticised as toothless. A broad one catching a large share of the packaged food supply invites the argument that it stigmatises affordable, shelf-stable food that many households depend on — and that argument carries real weight in a country where food insecurity is widespread.

How the definition resolves these four problems will determine whether it changes anything.

Public debate often collapses two separate claims: that ultra-processed foods are associated with poor health outcomes, and that processing itself causes those outcomes. The evidence for these is not equally strong, and being precise about the difference is what separates a credible article from a scare piece.

The association is well established. A large body of observational research links higher ultra-processed food consumption to increased risk of cardiovascular disease, type 2 diabetes, obesity, and all-cause mortality. These studies span multiple countries and cohorts, and the associations generally persist after adjusting for conventional dietary quality markers such as saturated fat, sugar and sodium. That persistence is the interesting part — it suggests something beyond nutrient composition is operating.

Causation is where the evidence thins. Observational studies cannot separate the food from the people eating it: ultra-processed food consumption correlates with income, time pressure, food environment, and a great deal else. The strongest causal evidence comes from controlled feeding trials, of which there are few, and those that exist are small and short.

The candidate mechanisms under investigation include energy density and the speed of eating, disruption of satiety signalling, effects of specific emulsifiers on the gut barrier and microbiome, and displacement — ultra-processed food crowding out whole food rather than causing harm directly.

Scale gives the question urgency regardless of mechanism. Ultra-processed foods account for roughly 60 percent of total daily energy intake for the average American adult, and the figure is higher for children. At that share of the diet, even a modest true effect matters at population level.

The United States is late to this, not early. Several countries already regulate on the basis of processing or nutritional profile, and their experience is the best available evidence for what the American definition might achieve.

Two patterns are worth extracting from this table.

First, most regulators have chosen nutrient thresholds over processing classification. Chile, Mexico, the UK and Australia all regulate on the basis of sugar, salt, fat and energy density, not on whether a food is industrially formulated. Nutrient thresholds are easier to write, easier to enforce, and harder to litigate. If the American definition ends up nutrient-based in practice, it will be following the majority.

Second, Chile's experience is the most studied and the most encouraging. Following the introduction of warning labels and marketing restrictions, researchers documented measurable reductions in purchases of products carrying warnings — with the marketing restrictions appearing to contribute meaningfully alongside the labels themselves. That is the closest thing to proof that this category of policy changes behaviour rather than merely informing it.

What no country has yet done is regulate ultra-processing as such, using a processing-based legal definition. That is what makes the American attempt genuinely novel, and it is also why it is taking so long.

A definition on its own changes nothing. What it enables is the point.

Front-of-package labelling is the most immediate application. The FDA's proposed front-of-package rule, issued in early 2025, focuses on added sugar and sodium. A federal ultra-processed definition creates the possibility of a processing-based label element as well — and the FDA's own 2026 agenda commits to summarising public comments and preparing options for a final front-of-package regulation.

Federal food procurement is the second lever, and possibly the larger one. School meals, military provisioning and federal nutrition assistance collectively represent enormous purchasing power. Procurement standards that reference a legal definition can shift manufacturer behaviour faster than consumer labelling does, because they change what gets bought at scale.

Marketing restrictions are the third. The FDA's 2026 agenda includes work with the Federal Trade Commission on industry guidelines limiting unhealthy food marketing to children. A definition gives such guidelines a boundary they currently lack — and the United Kingdom has already demonstrated the model, with advertising restrictions on less healthy food taking effect on 5 January 2026.

Research funding is the fourth and the slowest-acting. A standard definition makes studies comparable across institutions, which is a precondition for resolving the causation question that currently limits how far policy can defensibly go.

There is also a commercial consequence that will arrive before any of these. Once "ultra-processed" has a legal meaning, its absence becomes a marketing claim. Expect "not ultra-processed" positioning to appear quickly, and expect disputes about who may use it.

Understanding the opposition is not optional if you want to predict where the definition lands, and the industry case is more substantive than it is usually portrayed.

The central argument is that processing is not a hazard category. Pasteurisation is processing. Fortification is processing. Freezing is processing. Each of these has saved an enormous number of lives, and a framework that treats "processed" as inherently suspect risks communicating something false to the public. Industry submissions have pressed this point hard, and it is correct as far as it goes — NOVA's Group 1 explicitly includes pasteurised and frozen foods, but the public shorthand routinely loses that nuance.

The second argument concerns affordability and access. Shelf-stable, fortified, long-life foods are disproportionately what lower-income households buy, because they are cheap, transportable, and do not require refrigeration or rapid consumption. A regulatory framework that stigmatises this category without changing its price or availability may simply add guilt to constrained choices. In a country with significant food insecurity, this is a serious objection rather than a deflection.

The third is fortification specifically. Many ultra-processed products are vehicles for micronutrients — folic acid in flour, vitamin D in milk alternatives, iron in cereals. These fortification programmes have measurable public health results. A definition that flags fortified foods as undesirable sits awkwardly against decades of nutrition policy that deliberately created them.

The fourth is more procedural: that any definition drawn on "additives not used in home cooking" is arbitrary, because home cooking is not a fixed category and varies enormously by culture. Xanthan gum is a kitchen ingredient in some households and an industrial additive in others.

None of these arguments dissolve the epidemiological findings. What they do is constrain how broadly a defensible definition can be drawn, and they explain why the process has taken more than a year from Request for Information to OMB submission.

Advice to "eat less ultra-processed food" is useless without specifics, because the category covers a large share of what is actually available in a supermarket. These are substitutions that reduce Group 4 intake without requiring anyone to cook from scratch every evening.

Two things to note about this list. None of the replacements are expensive specialty products — oats, tinned tomatoes and plain yoghurt are among the cheapest items in any supermarket, which matters given the affordability objection raised above. And none require eliminating packaged food entirely, which is neither realistic nor necessary. The goal is shifting the proportion, not achieving purity.

The most efficient single change for most people is drinks. Sweetened beverages contribute a disproportionate share of ultra-processed intake, carry essentially no satiety value, and are the easiest category to substitute without affecting meals at all.

Three things are true simultaneously, and most coverage of this topic picks one and ignores the other two.

Ultra-processed food consumption is robustly associated with worse health outcomes across large populations, and the association survives adjustment for the nutrients that conventional dietary advice focuses on. That is a real finding and it is why regulators are moving.

The causal mechanism is not established, controlled trial evidence is thin, and the confounding in observational data is substantial. Anyone claiming the science is settled on causation is overstating it.

The definitional problem is genuinely hard, and a poorly drawn definition could produce perverse outcomes — penalising fortified foods that serve a nutritional purpose while exempting products that are simply traditional.

The practical guidance that survives all three caveats is unchanged and unglamorous: the more of your diet that comes from food you can recognise as food, the better, and the ingredient list is the only part of the packet obliged to tell you the truth about that. You can apply that today. The federal definition will make it enforceable, not newly true.