Is soya milk ultra-processed? Are almond drinks? Made plainly — water plus the bean or the nut, and nothing else — typically neither is. That is a minimally processed drink, sitting somewhere between group 1 and group 3 of the NOVA classification, depending on whether a little salt or oil has been added. The complication is that, from our own reading of UK shelves, most cartons carry more than that: a stabilising gum such as gellan, an emulsifier such as sunflower lecithin, added calcium and vitamins, and sometimes flavourings, sugar or a non-sugar sweetener on top. Which of those actually moves the product into ultra-processed territory is where the honest answer gets longer.
We have already been through this with oat drinks, and the reasoning carries across the chiller aisle. NOVA classifies products, not categories. Two soya drinks side by side on the same shelf can land in different groups.
What NOVA actually counts — and what it does not
NOVA sorts food by the extent and purpose of industrial processing. Group 1 is unprocessed or minimally processed food. Group 2 is culinary ingredients: sugar, salt, oils, butter, vinegar. Group 3 is processed food — group 1 foods with group 2 ingredients added, usually to preserve them or make them ready to eat. Group 4 is ultra-processed food: industrial formulations built from derived substances and cosmetic additives.
Two details do most of the work in a plant milk argument. The first is that group 2 ingredients do not create group 4. Adding salt or a little oil to a soya drink makes it processed, not ultra-processed. The second is that not every additive is a marker. Preservatives, antioxidants, stabilisers, gums, acidity regulators and added vitamins and minerals are group 3 territory under NOVA's own rules. The additives that signal group 4 are cosmetic ones — flavourings, flavour enhancers, colours, non-sugar sweeteners — along with industrial ingredients rarely used in a home kitchen, such as protein isolates, maltodextrin and modified starches.
Plain soya and almond drinks
The traditional method is old and simple. Soak soya beans, grind them with water, heat the mixture and strain out the solids. Almond drinks follow the same shape with almonds in place of beans, though almonds are expensive and the resulting drink is thin, which shapes what happens next in a factory.
A carton whose ingredient list reads water, soya beans, perhaps a pinch of salt, carries no ultra-processing markers on the additive-function reading. Nothing in soaking, grinding, heating and straining is industrial in NOVA's sense. We have found these drinks in UK supermarkets, and they typically sit in NOVA 1 to 3. We are not naming products, because manufacturers reformulate without announcement and the carton you bought last year is not evidence about the one on the shelf today.
What is usually on the carton
Beyond the plain versions, and again from our own scanning of UK shelves, a plant milk usually adds some combination of the following: gellan or another gum to stop the solids settling out, sunflower lecithin as an emulsifier, calcium carbonate and a vitamin blend for fortification, occasionally an acidity regulator, and in the sweetened and flavoured ranges, sugar, flavourings or a non-sugar sweetener.
Sort those by NOVA's own logic and they fall into two piles. The gums, the lecithin, the acidity regulator and the added calcium and vitamins are all group 3 entries on the additive-function reading — they keep the drink stable and nutritionally useful rather than making it taste or look like something it is not. The flavourings, the non-sugar sweeteners and any protein isolate are the group 4 markers. Sugar belongs in neither pile: it is a group 2 culinary ingredient, so a sugar-sweetened drink reads as group 3 on that count alone. On this reading, a plain fortified soya drink with a gum and a lecithin is typically a processed food, while a flavoured version, or one carrying a non-sugar sweetener, typically reads as ultra-processed. Check the label.
The part where coders disagree
That reading is defensible, but it is an interpretation, and other readings reach a different answer. The NOVA implementation used by Open Food Facts, for instance, commonly lands commercial plant drinks in group 4, treating the gum-and-emulsifier system as a cosmetic texture formulation rather than a simple stabiliser, and counting fortification as part of the picture of an industrially engineered product. That is the reading behind the widely repeated claim that plant milks are UPF, and it is the reading we applied to barista and standard oat drinks.
Neither position is obviously wrong. Braesco and colleagues put NOVA coding to the test in 2022, running two separate panels — 159 and 177 evaluators, working from two different food lists — and found Fleiss kappa values of 0.32 and 0.34, which the authors described as only fair agreement. In our own experience the arguments cluster around formulations like this one. Treat the group as a signal rather than a settled verdict, and read the ingredient list yourself.
Fortification is a nutrition question, not a processing one
Keep these two things apart. On the reading Cull applies, added calcium, iodine, B12 and vitamin D do not on their own move a product between NOVA groups. Separately from NOVA entirely, they matter: for anyone not drinking dairy, a fortified plant milk is a meaningful source of calcium, and unfortified minimal recipes usually skip it. A short ingredient list can therefore be the worse nutritional choice for the person choosing it. That is not a flaw in NOVA. It is a reminder that NOVA describes processing and says nothing about nutrients.
The other checks worth running are also independent of the NOVA group: added sugars in the sweetened ranges, and protein content, which soya drinks typically carry and almond drinks typically do not.
What to check on the carton
- Water and soya beans or almonds, and little else — minimally processed, NOVA 1 to 3.
- Gums, lecithin, acidity regulators— group 3 additives by function. Read alongside everything else on the list rather than as a verdict on their own.
- Flavourings, non-sugar sweeteners, protein isolates — the entries that typically point to NOVA 4.
- Added calcium and vitamins— not a processing marker, and usually worth having if the drink replaces dairy.
- Sugar— a group 2 culinary ingredient, so it points to group 3 rather than group 4. How much of it there is stays a separate nutrition check.
Or skip the memorising. Scanning the barcode in Cull's UPF scanner shows the NOVA verdict along with the additives and the nutrition, which is quicker than comparing five cartons of small print in a cold aisle.
Does a group 4 label mean the drink is bad for you?
No. Ultra-processed and unhealthy are correlated, not synonymous. Lane and colleagues pulled the observational evidence together in a 2024 BMJ umbrella review covering roughly 9.9 million participants, reporting associations between higher UPF intake and a range of health outcomes with varying certainty of evidence. Associations are not proof of cause, and that evidence is about UPF intake overall rather than about soya or almond drinks specifically.
The UK's own advisory committee has been careful here. SACN's 2023 position statement on processed foods and health, and its rapid evidence update published on 2 April 2025, both describe the associations as concerning while noting that limitations in the evidence make it unclear whether processing itself is the cause, rather than these foods simply tending to be high in energy, saturated fat, salt and free sugars. A plain fortified soya drink is not typically high in energy, saturated fat or salt, though sweetened and flavoured ranges can carry meaningful free sugars and coconut-based drinks carry saturated fat. Check the back-of-pack figures, which are a separate question from the NOVA group.
Sources
- Monteiro et al. (2016) — NOVA. The star shines bright, World Nutrition
- Braesco et al. (2022) — Ultra-processed foods: how functional is the NOVA system? European Journal of Clinical Nutrition
- Lane et al. (2024) — Ultra-processed food exposure and adverse health outcomes: umbrella review, BMJ
- Open Food Facts — NOVA groups for food processing
- SACN (2023) — position statement on processed foods and health, and SACN's rapid evidence update published 2 April 2025