Ideas developed from the author’s original article
From Broad Concepts to Operational Definitions
The debate around “ultra-processed foods” (“UPFs”) is entering a new phase. At the end of 2025, The Lancet published a 3-paper series on “UPFs”, covering health associations and policy recommendations1.
Much of the evidence underpinning this classification comes from observational studies that rely on dietary assessment methods not originally designed to assess food processing2.
While these studies identify associations between dietary patterns and health outcomes, they do not establish cause and effect or determine the independent contribution of specific ingredients to the observed associations.
One of the most significant developments from the Lancet paper series was the proposal of moving beyond the broad descriptive Nova classification by Monteiro et al.3 toward a more operational definition based on “markers of ultra-processing” (MUPs).
These MUPs tend to include ingredients and additives not generally used in home cooking. In practice, however, they also represent a subtle but important shift in how “UPF” is being conceptualised.
The discussion implicitly moves away from processing towards ingredients and formulation-level classification systems capable of being applied at product level and at scale.
This distinction matters because it means there are more questions to be answered such as which classification system is being applied, how is it operationalised, and what happens when it is used across real-world datasets.
These questions reflect a broader transition taking place across nutrition policy, where the conversation is increasingly focussed on data infrastructure, challenges, opportunities and implementation of health metrics.
The Lancet UPFs series1 expands upon the related literature by moving from a classification widely used in observational studies to a definition that can be applied in practice for policy purposes.
Yet, at the time of writing this article, these ingredient-level approaches have not been empirically tested against health outcomes.
Existing Policy Landscape Remains Primarily Nutrient-Based
Despite growing policy and scientific attention on food processing levels, most regulatory approaches for enabling healthier food choices (e.g. taxation, marketing restrictions, labelling) still operate primarily through nutrient-based thresholds.
© Copyright 2026 Global Food Research Program at UNC-Chapel Hill. Base map copyright © FreeVectorMaps.com
Front-of-pack warning labels in countries such as Chile and Mexico focus on nutrients of concern and, in some cases, sweeteners. Restrictions in the UK on products high in fat, salt, and/or sugar (HFSS) rely on a nutrient profiling model.
Similarly, the Nutri-Score in certain European countries evaluates products through nutrient composition for labelling. China requires sugar and saturated fat labelling and health warnings for children for mandatory implementation in 2027.
Although these systems do not classify foods according to their level of processing, food processing and ingredient-level definitions are now being considered in discussions around dietary guidance, marketing restrictions, taxation and/or front-of-pack labelling.
As these policy approaches use different underlying criteria, this raises an important question: what additional value does classifying foods by their level of processing provide beyond existing nutrient-based systems?
The Rise of Ingredient-level Classification
The proposed MUPs approach attempts to create a more scalable and standardised way of identifying “UPFs” using ingredient list data. In theory, this allows classification to be applied consistently across large manufacturer and retailer datasets, but it also raises unresolved questions within the food processing debate.
One question is whether regulation is ultimately targeting specific ingredients or foods as a broader category. Discussions often move interchangeably between emulsifiers, stabilisers, sweeteners, flavourings, industrial formulations and processed foods themselves, despite these not necessarily representing equivalent policy targets.
Another challenge relates to evidence translation based on the available science to date. As a result, it leaves a gap between the descriptive categories used in dietary surveys and ingredient-level systems proposed for regulatory implementation.
The growing shift toward ingredient-level operationalisation therefore changes the nature of the debate. At the same time, implementation remains inherently difficult.
Currently, there is no globally harmonised MUP framework, raising practical questions around who determines the official list of markers, how lists are updated and how disagreements are resolved as (re)formulation practices evolve.
Emerging Attempts to Bridge the Implementation Gap
In May 2026, Healthy Eating Research (HER) convened an expert panel to evaluate existing UPF definitions and approaches for guiding policy development4. The report acknowledged that processing alone is not sufficient to determine the overall healthfulness of foods and instead is as an additional lens that may complement existing dietary assessment approaches.
In addition, inclusion of specific markers is not necessarily because those ingredients are individually harmful, but because they are viewed as indicators of foods categorised as “ultra-processed”.
However, while broader dietary patterns classified as “UPFs” may have been associated with adverse health outcomes in observational studies, it remains unclear whether the proposed ingredient-level markers used for policy capture those same associations.
The HER panel brought practical implementation challenges into sharper focus. These included maintaining and updating ingredient databases within rapidly evolving food supply chains, ensuring sufficient transparency around formulation practices, and managing the reality that static marker lists may quickly become outdated.
At the same time, the project reflects expansion of the field by applying proposed metrics to more than 90,000 products, highlighting how food classification systems need to consider real-world data to understand policy implications.
Reformulation and Shifting Incentive Structures
One of the most important implications of ingredient-level approaches relates to reformulation incentives. The Lancet series critiques nutrient reformulation by arguing that reductions in fat, salt and/or sugar may increase the use of additives or industrial ingredients.
Yet ingredient-level operationalisation could itself create a different set of reformulation incentives, which may take the form of optimisation away from “less sugar” towards “less emulsifier”, expanding previous trends on “clean(er) label” or “natural” ingredient substitution strategies.
If ingredient-level criteria become the primary regulatory gatekeeper, reformulation may lead to prioritisation of additive substitution and formulation optics over measurable nutritional improvements.
This creates a broader policy question: what exactly is regulation trying to incentivise – nutritional composition, degree of processing, category restrictions, ingredient familiarity, formulation simplicity, eating behaviour, or some combination of these factors?
California as an Emerging Real-World Test Case
A number of these tensions are now beginning to materialise in regulatory approaches.
California has become one of the clearest examples of attempts to operationalise the level of food processing within legislation. In 2025, the state introduced a formal “UPFs” definition for school food restrictions through AB 1264. More recently, AB 2244 proposed applying this classification through a certification scheme for non-“UPFs” and preferential retail placement for certified products.
Rather than relying solely on Nova categorisation, California’s approach combines ingredient-level markers with nutrient thresholds. Products may be classified as “UPFs” if they contain one or more listed additives while also exceeding thresholds for saturated fat, added sugar or sodium. However, products containing non-nutritive sweeteners or polyols may still fall within the definition even where nutrient thresholds are not exceeded.
The result is a system where HFSS does not always equal “UPFs”, and vice versa, which creates potential divergence. Nutritionally poor products may avoid “UPF” classification through “natural” substitutions, while products that have improved nutritionally may remain within restricted categories because of ingredient composition.
Alongside policy-led approaches, a growing number of third-party certification schemes have also been developed, particularly in the US, to certify foods as non-“UPF”. These schemes often apply different criteria and thresholds, further illustrating the fragmentation currently emerging within operational definitions.
Product-Level Data is Becoming Central
One of the clearest changes emerging is that the debate is moving beyond a single classification and increasingly overlapping with wider discussions on data infrastructure.
Policy impact ultimately depends not on whether a classification system appears coherent theoretically, but on what happens when it is applied across real-world products, retail environments, purchasing and consumption datasets.
A framework may appear conceptually robust yet generate unexpected divergence, inconsistencies or implementation challenges once operationalised at scale.
Questions around ingredient transparency, data access, and successful implementation are therefore becoming central to the future of food and nutrition policy.
Final Reflections
A key challenge for “UPFs” moving forward is considering where the concept genuinely adds value beyond existing policy approaches. Most stakeholders working in this space ultimately share similar goals: supporting healthier and more sustainable diets.
However, before “UPF” becomes embedded within regulatory frameworks, there needs to be greater clarity around what the concept is intended to achieve and whether current operational approaches remain coherent when applied consistently across real-world food systems and datasets.
Whatever direction the conversation takes, the future of “UPFs” feels likely to become more data-led, not only “definition-led”. As such, the next chapter may depend less on continuing debates around classification itself, and more on understanding what happens when these approaches are applied in practice.
Without this, there is a risk that unintended consequences and potential divergence beyond existing nutrient-based approaches remain poorly understood, particularly as regulatory development in this area continues to accelerate.
