BEFORE WE TELL PEOPLE TO AVOID ULTRA-PROCESSED FOODS, CAN THEY IDENTIFY THEM?

As concern about ultra-processed foods (UPFs) grows, policy makers and activists are increasingly discussing measures such as front-of-pack UPF warning labels. The logic is appealing: if consumers struggle to identify UPFs, simply label them and the problem is solved.

In my recent Public Health Nutrition commentary (1), I argue that the issue is more complicated than that. Before recommending that people avoid a category of foods, we should first ask a basic question: can ordinary consumers reliably identify the foods that belong in that category?

Evidence from several countries suggests they cannot. Importantly, I argue that this should not be interpreted as a failure of consumer knowledge or nutrition literacy. Instead, it reflects a mismatch between how the NOVA classification system works and how people actually make food choices.

Three reasons the problem is built into the instrument

The first problem is perceptual accessibility.

NOVA defines UPFs largely by features such as cosmetic additives and aspects of industrial formulation. These are not things consumers can easily observe. A shopper can see packaging, branding, ingredients and nutrition information, but whether an additive is serving a “cosmetic” rather than a technical purpose (e.g., a preservative) is often unclear. Understanding such distinctions may require specialist knowledge that most consumers do not possess.

The second problem is schema compatibility.

When people shop for food, they do not analyse products using NOVA’s classification rules. Instead, they use intuitive mental shortcuts. They judge foods according to whether they appear natural, familiar, homemade, nutritious or “heavily processed”. Research shows these heuristics are coherent and predictable, but they frequently do not align with NOVA criteria. For example, many consumers assume that a fortified breakfast cereal or yoghurt is less processed because it seems healthy, even when NOVA classifies it as a UPF.

This means consumers are not ignoring food processing. They are thinking about it, just not in the same way that NOVA conceptualises it.

The third problem is ecological validity.

The same type of food can move between NOVA categories depending on the brand and formulation. One bread may be classified as Group 3 (Processed foods), while another almost identical bread may be classified as Group 4 (Ultra-processed foods) because of the presence of a particular emulsifier. Likewise, one yoghurt may be minimally processed while another is classified as ultra-processed because of added flavourings or colours.

For consumers, this creates an unstable system. Learning the classification of one product does not reliably help with the next purchasing decision.

Why more education is the wrong lever

A common response is to argue that consumers simply need more education about UPFs.

My commentary argues that this response misunderstands the nature of the problem. Consumers already rely on well-established cognitive shortcuts when selecting foods. These shortcuts are shaped by decades of dietary guidance, food marketing and everyday experience. They operate automatically and efficiently in the time-pressured environment of supermarkets. Simply providing more information is unlikely to replace them.

When the obstacle is built into the design of the classification system itself, education becomes a weak policy lever.

Would UPF warning labels help?

UPF warning labels would certainly make identification easier. Consumers would no longer need to determine for themselves whether a food is a UPF.
However, identification is only part of the challenge.

The fourth cognitive validity criterion I discuss is subcategory discriminability: whether a classification system meaningfully distinguishes foods with different nutritional implications. Here, NOVA faces another difficulty. The UPF category includes both energy-dense, nutrient-poor foods such as sugar-sweetened soft drinks and confectionery, and foods such as some wholegrain breads, fortified breakfast cereals and fortified plant-based milks (e.g., soy milks) that can contribute important nutrients to the diet.

A warning label might tell consumers that all of these products are Ultra-processed according to NOVA. It does not tell them how these foods should be treated from a nutritional perspective.

This is why I argue that easier identification alone cannot resolve concerns about the usefulness of the UPF category itself.

The bottom line

My commentary does not argue that concerns about UPFs are unfounded, nor does it suggest that all UPFs are healthy. Rather, it argues that consumer-facing food classifications should be evaluated according to whether ordinary people can understand and apply them in practice.

Before building dietary guidance and warning labels around a classification system, we should ensure that the system is perceptible, intuitive, consistent and capable of distinguishing between foods that differ meaningfully in nutritional value. Until then, public health nutrition may be better served by guidance based on characteristics consumers can readily observe and use.

Read more:

1. Louie JCY. Before we tell people to avoid ultra-processed foods, can they identify them? The case for cognitive validity criteria in food classification. Public Health Nutr, 2026.