Nutrition
Is Pistachio Milk Ultra-Processed? 37 US Cartons Against the Definition
We applied the published test to 37 US cartons. Most of the plant shelf fails it on a single gum, and two of the three cartons that pass are pistachio milk.
Quick answer
It depends entirely on the carton, and the honest answer has a number in it. Applying the practical test from the NOVA system to the declared ingredients of 29 US plant milks, 26 declare at least one marker of ultra-processing and 3 declare none. Two of those three are pistachio milk: Elmhurst 1925's three-ingredient carton and Three Trees' four-ingredient carton. A third pistachio carton declares three gums and added pistachio flavor. None of the 8 cow's milk cartons we read declares a marker. The United States has no official definition yet: a white paper proposing one reached the White House review office on 3 August 2026 and was still pending on 20 November 2026.
It depends on the carton, and on this shelf that is a bigger difference than it sounds. Applying the published test to the declared ingredients of 29 US plant milks, 26 declare at least one marker of ultra-processing and 3 declare none. Two of those three are pistachio milk. None of the 8 cow's milk cartons alongside them declares a marker at all.
We said on our plant milk ingredients audit that we describe labels rather than assigning products to a processing category. This page is the other half of that. It reports what happens when you run a real shelf through the definition people are arguing about, and what the result does and does not tell you. It passes no judgment on whether the category is good for you.
What test are we applying?
The one the researchers who coined the term published. Monteiro and colleagues set it out in Public Health Nutrition in 2019, and it is deliberately short enough for a shopper to use:
A practical way to identify an ultra-processed product is to check to see if its list of ingredients contains at least one item characteristic of the NOVA ultra-processed food group, which is to say, either food substances never or rarely used in kitchens (such as high-fructose corn syrup, hydrogenated or interesterified oils, and hydrolysed proteins), or classes of additives designed to make the final product palatable or more appealing (such as flavours, flavour enhancers, colours, emulsifiers, emulsifying salts, sweeteners, thickeners, and anti-foaming, bulking, carbonating, foaming, gelling and glazing agents).
Two halves, then. Substances you would not find in a kitchen, and what the paper elsewhere calls cosmetic additives. Read the second list again and notice what is on it: thickeners and gelling agents. Every gum on a plant milk carton is one of those.
Notice what is absent, too. Added vitamins and minerals appear nowhere in the marker list. Neither do plant oils, which the paper puts in a different group and says plainly "are also not ultra-processed", along with sugar and salt.
How does the shelf sort?
Almost all one way. Here is every base in our 37-carton set, with how many cartons declare a marker.
| Base | Cartons | Declare a marker | Declare none |
|---|---|---|---|
| Pistachio | 3 | 1 | 2 |
| Rice | 2 | 1 | 1 |
| Almond | 4 | 4 | 0 |
| Cashew | 4 | 4 | 0 |
| Coconut | 3 | 3 | 0 |
| Soy | 4 | 4 | 0 |
| Oat | 4 | 4 | 0 |
| Macadamia | 2 | 2 | 0 |
| Walnut | 1 | 1 | 0 |
| Pea | 1 | 1 | 0 |
| Hemp | 1 | 1 | 0 |
| All plant milks | 29 | 26 | 3 |
| Cow's milk | 8 | 0 | 8 |
Counted by class rather than by carton, the plant shelf runs: thickeners on 22 of 29, a flavor declared as a class on 15, an emulsifying salt on 11, an emulsifier on 8. Two more cartons declare vanilla extract, which we did not count, for reasons in the methodology. Not one carton in the set declares a color or a sweetener of any kind.
Which cartons declare nothing?
Three, out of twenty-nine, and the list is short enough to print in full.
- Elmhurst 1925 Unsweetened Pistachio Milk. Filtered water, pistachios, Himalayan pink salt.
- Three Trees Organic Unsweetened Pistachio Nutmilk. Filtered water, organic pistachios, organic almonds, French grey salt.
- Hain Celestial Rice Drink Original. Filtered water, organic brown rice (partially milled), organic expeller pressed canola oil and/or organic safflower oil and/or organic sunflower oil, tricalcium phosphate, sea salt, vitamin A palmitate, vitamin D2, vitamin B12.
A fourth sits on the line and we have put it on the marker side. Glencourt's Original Organic Ricemilk declares no cosmetic additive either, but its second ingredient is "organic partially milled hydrolyzed rice", and a starch hydrolysate belongs with the maltodextrin and dextrose the paper names in its first half. It is the only row in the set whose answer turns on a single reading, which is why we flagged it rather than buried it. It also declares 6 g of added sugars a cup on a label that lists no sweetener, which is what hydrolysis does and the best evidence that the word is doing real work.
Read that as a fact about three cartons rather than about the nut. The same base supplies one of the most heavily marked products in the set: the Elmhurst Harvest record declares xanthan gum, guar gum, gellan gum and pistachio flavor with other natural flavors. A pistachio drink that adds pistachio flavor is as far into group 4 as this shelf goes.
Why does one gum settle it?
Because the test is a threshold, and it was built to be.
Wegmans sells an organic unsweetened almondmilk whose entire ingredients list is organic almondmilk (filtered water, organic almonds), sea salt, gellan gum. Four ingredients, no fortificant, no oil, no flavor. It is ultra-processed under the test, and it sorts into exactly the same group as a sixteen-ingredient cashewmilk carrying canola oil, three gums, natural flavor and an eight-part vitamin and mineral blend.
Forager Project's unsweetened cashewmilk gets there on konjac. Oatly's Original gets there on dipotassium phosphate alone, with no gum anywhere on the label. Three of the four oat cartons we read have no thickener at all and still land in group 4, on a buffering salt added so the drink survives hot coffee.
So the test answers one question, cleanly: does this carton contain anything an industrial kitchen uses and a home kitchen does not? It does not rank. A shopper who wants to know which of two cartons is more processed gets no help from it, because 26 of 29 give the same answer.
What can the test not see?
The two things that most change what a plant milk does for you.
Fortification is invisible to it. Calcium carbonate is not a marker, so the carton declaring 451 mg of calcium a cup and the carton declaring 0 mg sort identically. Our calcium figures show that gap running across the same shelf, and it is the single largest nutritional difference in the whole dataset.
Added oil is invisible to it as well, because the paper classes plant oils as culinary ingredients. Oatly declares low erucic acid rapeseed oil and Wegmans' oat declares sunflower oil, and neither counts. The carton gets its group 4 label from a phosphate salt declared inside its "contains 2% or less of" group.
There is a sharper version of that point. Two of the three cartons declaring no marker are also the two declaring the least nutrition of anything on the shelf. Elmhurst 1925 publishes calories and protein and no other figure. Three Trees declares 45 mg of calcium. Neither carries an added vitamin or mineral at all. The third, Hain's rice drink, is fortified to 300 mg and sits at 1 g of protein, so a clean result on the processing test says nothing consistent about the panel either way. We found the same pattern from the other direction when we read the ingredients across all 37 cartons and when we compiled the panel dataset.
Is there an official US definition?
No, and the agencies that are writing one say so themselves. From the Federal Register notice of 25 July 2025, in which FDA and USDA asked the public for help:
There is no single, universally accepted definition of UPFs, and the definition of such foods has varied considerably over time.
The same notice names NOVA as "the most common classification" and describes how it identifies group 4 "based on multiple factors; these factors include things like the use of certain ingredients and substances (such as emulsifiers, bulking agents, or thickeners), industrial processing technologies, as well as sophisticated packaging".
Where that effort stands, as of 20 November 2026: the Federal Register holds two documents from the agencies on the subject, both from 2025, the request for information and a 19 September extension of its comment period. No definition has been published. A document titled "White Paper: Proposed Definition of Ultra-Processed Food", RIN 0910-ZD60, from HHS and FDA, reached the Office of Information and Regulatory Affairs on 3 August 2026 and was still listed as pending review when we checked, 109 days later. Its stage is recorded as Notice, with no legal deadline, and it is not classed as economically significant.
How differently would a state bill sort the same cartons?
Very differently, and the federal notice supplies the evidence. It lists three shapes of definition proposed by US states, with nine bills cited by number. Run our 29 cartons through each shape and the answers barely overlap.
| Shape of definition | Example cited in the notice | What it would catch here |
|---|---|---|
| Substances intended to have an effect on the food: stabilizers and thickeners, coloring or flavoring agents | Pennsylvania H.B. 1132; California A.B. 1264 | close to the NOVA result, 25 of 29, since it names the same functional classes |
| Foods that have undergone certain processing steps, such as hydrogenation of oils or hydrolysis of proteins | Massachusetts H.B. 539 | almost nothing. No carton declares a hydrogenated oil or a hydrolyzed protein. One declares hydrolyzed rice, which is a starch |
| Foods containing one of a fixed list of 10 to 15 named ingredients | Florida S.B. 1826 (11); Louisiana S.B. 117 (15); North Carolina H.B. 874 (11); Arkansas H.B. 1962 (10); Alabama H.B. 580 (11); South Carolina S.B. 589 (11); Kentucky H.B. 439 (11) | depends on which 11, and the bills do not agree on them |
That is the practical problem under the argument. The same 29 cartons are nearly all ultra-processed, almost none ultra-processed, or anybody's guess, and nothing about the cartons changed.
What does US law actually test on a carton?
Nutrients, and nothing about additives. FDA's final rule redefining the word "healthy" was published on 27 December 2024, took effect on 25 February 2025 and has a compliance date of 25 February 2028. The codified criteria at 21 CFR 101.65(d)(3)(ii)(A) set four conditions for a dairy product:
| Condition | Requirement for a dairy product |
|---|---|
| Food group equivalent | at least 2/3 cup equivalent dairy |
| Added sugars | no more than 5% DV |
| Sodium | no more than 10% DV |
| Saturated fat | no more than 10% DV |
Plant milks are judged there. FDA says in the rule that plant-based milk alternatives "would be evaluated against the criteria for the dairy food group", and that "it is only those plant-based dairy alternatives that have similar nutrition composition to dairy that will be included in the dairy group" for this purpose. The rule describes the products it had in mind as ones that "provide similar amounts of protein, calcium, potassium, vitamin D, and other nutrients".
Put the two frameworks side by side and they disagree about this shelf. The test researchers use rewards a short label. The only test written into US law rewards a nutrient profile close to milk's, which on a plant carton usually means fortification, which means a longer label. A three-ingredient pistachio milk declaring 3 g of protein and publishing no calcium figure is the cleanest carton we read on one test and a weak candidate on the other.
Does any of this make pistachio milk bad for you?
The classification describes how a food was manufactured. The research behind it looks at what whole diets do over years, which is a different question from what one carton does for you. FDA and USDA put the difficulty better than we could, in the same 2025 notice:
On the other hand, foods considered to be ultra-processed may also include foods such as whole grain products or yogurt, which are known to have beneficial effects on health and are recommended as part of healthy dietary patterns. It is important therefore to consider unintended consequences of an overly-inclusive definition of UPFs that could discourage intake of potentially beneficial foods.
A fortified almondmilk with gellan gum in it is ultra-processed and supplies 451 mg of calcium. An unfortified pistachio milk is not ultra-processed and supplies almost none. Deciding between them on the classification alone gets you the wrong carton.
What we would actually tell a reader: if a short ingredients list matters to you, pistachio is the base with the shortest labels on this shelf, and which carton to buy is a real choice rather than a marketing one. If the drink is doing nutritional work in your diet, read the panel instead, because protein and calcium are where the cartons differ most and neither shows up in the processing test.
Methodology
What we classified. The complete declared ingredients list for 37 cartons sold in the United States: 29 plant milks across eleven bases and 8 cow's milks, two each of whole, 2% reduced fat, 1% lowfat and fat-free. These are the same 37 cartons as our ingredients audit and our nutrition panel dataset, so three datasets now describe one set of products.
The test we applied, and only that test. The practical identification rule quoted above, from Monteiro et al. 2019. We classified each carton on the ingredients its label declares, against the marker classes that paper names. We did not apply any other system, did not score degrees of processing, and did not assign NOVA groups 1 to 3 to anything, because the practical rule only answers one question.
The judgment calls, and whether they change anything. Six were needed. Four of them change no carton's answer, which is worth knowing before anyone argues about them. One changes ten rows and we have set it out at length.
- Gums, starch thickeners and carrageenan count as thickeners or gelling agents, both named classes.
- Lecithins count as emulsifiers. The paper names "soya lecithin as emulsifier" as an example.
- Phosphate and citrate salts not presented as fortificants count as emulsifying salts: dipotassium phosphate, disodium phosphate, sodium citrate, potassium citrate. Reading the citrates as acidity regulators instead would change no carton's answer, because all three cartons declaring one also declare a thickener.
- Pea protein counts as a substance of no or rare culinary use, by analogy with the soya protein isolate the paper names. All three cartons declaring it also declare a flavor and a gum, so this call changes nothing either.
- Vanilla extract and fenugreek do not count, because they are kitchen ingredients declared by name rather than a flavoring declared as a class. Both cartons involved declare a thickener anyway.
- Added vitamins and minerals do not count, because no vitamin or mineral appears in the marker list and the paper states separately that pasteurized milk is not ultra-processed. This is the call that decides the dairy result. Counting fortificants as markers would flip both rice drinks and all eight cow's milks, leaving two cartons in the whole set of 37 on the clean side. We read the published rule as clear on this, and anyone who disagrees can recount it from the CSV, which records the fortificants per row.
The one carton whose answer rests on reading a single ingredient is Glencourt's rice drink, discussed above.
Where the lists came from. Thirty-four are branded records in USDA FoodData Central, each carrying its FDC ID in the CSV. Three come from brand product pages: Three Trees and Pacific Foods, and Elmhurst 1925 carried from our barista label audit of 12 October 2026. Classification was done on 20 November 2026 against the ingredient strings collected on 16 November 2026.
Limitations. This reads labels, not factories. A classification built on declared ingredients cannot see a processing step that leaves nothing to declare, which is the Massachusetts-shaped objection and a real one: a drink can be made by high-pressure homogenization, enzyme treatment and ultra-high-temperature sterilization and still declare three ingredients. FoodData Central records are submitted by label owners and are not audited, and in this set they run from February 2018 to October 2023, so a record may not match a carton on a shelf today. Sample sizes per base are one to four cartons. Store brands are over-represented. And the result of the test is not a nutritional judgment, which is the whole point of the section above it.
Full dataset, with every marker class recorded per carton: plant-milk-upf-classification-2026-11.csv.
Reuse and citation
This dataset is published under CC BY 4.0. Use it, including commercially, with attribution.
Suggested citation: Ricci, E. (2026). Ultra-processed by whose definition: the NOVA identification test applied to 29 US plant milks and 8 cow's milks. pistachiomilk.org. https://pistachiomilk.org/is-pistachio-milk-ultra-processed/
Version history
- v1.0 (20 November 2026). First version. The NOVA practical identification test applied to the 37 cartons of our ingredients audit: 26 of 29 plant cartons declare at least one marker, 3 declare none, 2 of those 3 are pistachio, and 0 of 8 cow's milk cartons declare one. Federal status checked against the Federal Register and the OIRA pending review list on 20 November 2026. CSV, methodology and this history published with it.
Frequently asked questions
Is pistachio milk ultra-processed?
Does one gum make a plant milk ultra-processed?
Is cow's milk ultra-processed?
Has the US government defined ultra-processed food?
Does ultra-processed mean unhealthy?
Allergy note: Pistachios are a tree nut. If you have a nut allergy, avoid pistachio milk and pistachio products, and check labels for cross-contamination warnings. More on pistachio allergy, including who has to declare it. This article is general information, not personalized dietary or medical advice.