NOVA is the classification system behind the term ultra-processed food. It was developed by Carlos Monteiro and colleagues at the University of São Paulo, refined through the 2010s, and adopted in a 2019 report for the UN Food and Agriculture Organization. Its central idea is simple and slightly radical: sort foods not by their nutrients but by the extent and purpose of the industrial processing behind them. A food's calories, sugar and salt tell you one story. NOVA asks a different question — what was done to this food, and why?
The answer places every food in one of four groups. The groups are categories, not a graded scale — a food is in group 4 or it is not, and counting additives does not move it further up some ladder.
The four NOVA groups
Group 1 — unprocessed and minimally processed foods
Foods altered only by processes like removing inedible parts, drying, crushing, pasteurisation, freezing or non-alcoholic fermentation, with nothing added. Fresh and frozen fruit and vegetables, grains, plain flour, eggs, milk, plain yoghurt, meat, fish, nuts, dried herbs, tea and coffee all sit here. Freezing peas does not change what they are.
Group 2 — processed culinary ingredients
Substances extracted from group 1 foods or from nature, used to season and cook: salt, sugar, honey, butter, plant oils, vinegar. You rarely eat these on their own — they exist to turn group 1 foods into meals.
Group 3 — processed foods
Made by adding group 2 ingredients to group 1 foods, chiefly to increase durability or enhance flavour, and usually recognisable from two or three ingredients. Canned vegetables, legumes and fish, salted or smoked meats, fruits in syrup, freshly made unpackaged bread and traditional block cheese are the classic examples. This is most of what humans have called food preservation for centuries.
Group 4 — ultra-processed foods
Industrial formulations made mostly or wholly from substances derived from foods, plus additives, with little or no intact whole food, produced through a sequence of industrial processes. Soft drinks, packaged snacks, mass-produced sliced bread, most breakfast cereals, reconstituted meat products, instant noodles, many ready meals and confectionery are the textbook cases. If the ingredient list reads like a formulation rather than a recipe, this is the group it usually belongs to.
Notice that the same everyday food can appear in different groups depending on how it was made. Bread is minimally processed at a bakery counter and typically NOVA 4 in a supermarket bag. Traditional block cheese is group 3; processed slices are group 4. Plain yoghurt is group 1; the flavoured fruit-corner pot usually is not. The category belongs to the product, not the food — which is why the only honest answer to “is X ultra-processed?” is usually “check the label”.
Where the line between group 3 and group 4 actually falls
This is the part most coverage gets wrong. The line is not “contains additives”. Under NOVA, a food is group 4 when it contains markers of ultra-processing: ingredients never or rarely used in home kitchens, or classes of additives whose purpose is to imitate or enhance sensory qualities, or to disguise undesirable ones, in the final product.
The markers come in two families:
- Food substances of no or rare culinary use— high-fructose corn syrup, invert sugar, maltodextrin, hydrogenated or interesterified oils, hydrolysed proteins, protein isolates, mechanically separated meat. Nobody keeps these in a kitchen cupboard.
- Cosmetic additives— flavours and flavour enhancers, colours, non-sugar sweeteners, emulsifiers, thickeners, gelling agents, anti-foaming, bulking and glazing agents. Their job is to make the product look, taste or feel like something it otherwise would not.
The crucial nuance: additives used purely to preserve — preservatives, antioxidants and stabilisers used for durability or to resist microbial contamination — do not by themselves make a food ultra-processed. Monteiro and the FAO explicitly place fruits in syrup with added antioxidants and dried salted meats with added preservatives in group 3. So a single preservative on a label is not proof of NOVA 4. An emulsifier, a colour, a non-sugar sweetener, a flavouring or a protein isolate typically is. One marker is enough; two markers do not make a food more ultra-processed than one.
If you want the group 3 versus group 4 distinction worked through with UK examples on each side, we cover it in UPF vs processed food and turn it into a shelf-ready checklist in how to spot ultra-processed food on the label.
What the evidence says — honestly
The research linking ultra-processed food to poor health is large, consistent and overwhelmingly observational. That last word matters.
The biggest summary is Lane et al., published in the BMJ in 2024: an umbrella review covering 45 pooled meta-analyses and around 9.9 million participants. Higher UPF exposure was consistently associated with 32 adverse health outcomes, with the strongest evidence grades for cardiovascular-disease mortality, type 2 diabetes, common mental disorders and all-cause mortality. These are associations — people who eat more UPF fare worse — not proof that ultra-processing itself is the cause.
The one genuine causal signal comes from Hall et al., 2019, in Cell Metabolism: a randomised controlled trial in which 20 inpatients spent two weeks each on ultra-processed and unprocessed diets matched for presented calories, sugar, fat, sodium, fibre and macronutrients. On the ultra-processed diet, people ate about 500 kcal a day more and gained weight. It is the strongest evidence that ultra-processing itself drives overeating — and it is one small, short-term study, so it should not be overstated.
The Lancet's 2025 three-part UPF series, reviewing 104 papers, reached the same measured verdict: the disease link is consistently observed across cohorts but remains associative, and the causal chain is incomplete. The candidate mechanisms — energy density, low phytochemicals, a disrupted food matrix, additives — are plausible and unproven.
How reliable is NOVA?
Less than its popularity suggests. In Braesco et al., 2022, more than 150 experts classified the same set of foods and reached only fair agreement — a Fleiss kappa of about 0.32 for marketed foods and 0.34 for generic ones. The same product can land in different groups depending on who codes it and how much ingredient detail they can see. NOVA has serious academic critics, and this is their strongest card.
A second limit is just as important: ultra-processed and unhealthy are correlated, not synonymous. A wholemeal high-fibre sliced loaf can be NOVA 4 and still be a better choice than a white bakery loaf that is not. Classic canned baked beans are often group 3. A fortified plant milk can carry a NOVA 4 label and still be nutritionally useful. NOVA is one lens on food, not a verdict.
What the UK's official position is
Cautious, and worth representing accurately. SACN — the UK's Scientific Advisory Committee on Nutrition — published its position statement on processed foods and health in July 2023. Of eight processing-classification systems reviewed, NOVA was the only one it judged potentially suitable for UK use, but SACN flagged concerns about its practical application and consistency, and said the largely observational evidence should be treated with caution. It did not conclude that UPF is harmful independently of nutrient content.
SACN's April 2025 rapid evidence update held the same line: the associations remain of concern but carry significant limitations, and “on balance, most people are likely to benefit from reducing consumption of processed foods high in energy, saturated fat, salt and free sugars and low in fibre.” There is no UPF-specific NHS guidance; official advice still runs through the Eatwell Guide. Meanwhile UK adults get roughly 50 to 57 per cent of their daily energy from ultra-processed food — among the highest shares in Europe — which is exactly why a working understanding of NOVA is more useful here than almost anywhere else.
How Cull applies NOVA
Reading all of the above off an ingredient list in a supermarket aisle is a lot to ask. Cull does it per scan. Point the app at a product and it shows the NOVA group alongside the health score, and treats processing level as one of three scoring factors — next to nutritional quality and additive safety, never instead of them. That mirrors the science: processing matters, and so does what is actually in the food. A NOVA 4 wholemeal loaf and a NOVA 4 soft drink do not get the same score, because they are not the same food.
The NOVA group, the score and the reasoning behind both are free. You can read more about the scanning side on our UPF scanner page.