Nutri-Score vs Traffic Lights vs Stars: Food Labels Compared

Buy the same biscuit in five countries and the front of the packet will tell you five different things about it.

In Chile it may carry a black octagon reading HIGH IN SUGAR. In France, a letter grade from A to E. In the United Kingdom, three coloured circles for fat, sugar and salt. In Australia, a star rating out of five. In the United States, quite possibly nothing at all — though that is changing.

These systems were designed independently, by different regulators, using different logic, to solve the same problem: the nutrition panel on the back of the packet is accurate, comprehensive, and almost nobody reads it while shopping. Front-of-pack labelling is the attempt to compress that information into something a person will actually look at in the two seconds they spend choosing between two products.

This guide explains how each major system works, what happens when the same product is run through all of them, which designs have evidence behind them, and what is changing in 2026 — including the United States preparing a final front-of-package regulation and Australia moving its star rating from voluntary to mandatory.

Every front-of-pack scheme in use today is a variation on one of two ideas: summarise everything into one symbol, or flag individual problem nutrients.

The split between summary scores and nutrient-specific flags is the most consequential design decision in this entire field, and it determines almost everything about how a system behaves.

Summary systems — Nutri-Score, Health Star Rating — compute a single figure by balancing beneficial components against detrimental ones. They are fast to read and easy to compare. Their weakness is offsetting: a product high in sugar can score acceptably if it also contains fibre or protein, and the resulting symbol conceals which is which.

Nutrient-specific systems — traffic lights, warning octagons — report each nutrient separately against thresholds. They cannot be gamed by offsetting, because a high sugar figure produces a red light or a warning regardless of what else is present. Their weakness is cognitive load: three or four separate signals require more interpretation than one.

Warning systems are the most aggressive variant. They display nothing at all when a product is within thresholds, so the label is purely a negative signal. There is no reward for scoring well, only a penalty for scoring badly.

The clearest way to see how these differ is to run comparable products through each logic. The following are illustrative of how the systems behave rather than certified scores for specific branded products.

Three patterns emerge, and each has driven genuine political conflict.

Whole foods that are fat-dense score poorly on summary systems. Olive oil and nuts are the standard examples, and they are the basis of the most persistent objection to Nutri-Score — that a system rating extra virgin olive oil below a diet cola misrepresents nutritional reality. Defenders respond that the score is per 100g and oils are not consumed in 100g quantities. Both points are fair, and the dispute has had real political consequences in southern Europe.

Offsetting is visible in every summary row. The cereal scores respectably because fibre and protein counteract sugar. Under Chile's warning system the same product simply reads HIGH IN SUGAR, because there is nothing to offset against.

Artificially sweetened drinks expose a different gap. They contain almost nothing the algorithms measure, so they score well on nutrient-based systems despite being ultra-processed by any reasonable definition. No major front-of-pack system currently captures processing, which is precisely the gap the American ultra-processed food definition work may eventually address.

Nutri-Score is the most widely debated of these systems and the least understood, so it is worth opening the box.

The algorithm assigns negative points for components to limit and positive points for components to encourage, then subtracts one from the other and maps the result onto a letter from A to E.

Several design details explain the results people find surprising.

The protein cap exists to prevent high-protein, high-salt products such as processed meat from scoring well purely on protein content. Without it, cured meats would grade far better than dietary guidance supports.

Separate scales apply to fats and oils, to cheese, and to beverages. This is the same category-adjustment logic used by the Health Star Rating and for the same reason: a single scale across all foods produces nonsense comparisons.

Total sugars is used rather than added sugars, which is why plain milk and plain yoghurt incur sugar points for naturally occurring lactose. This is a known limitation, and it recurs across almost every nutrient-based system because added-sugar data is not universally declared.

And the score is per 100g regardless of how much is actually consumed. This is the origin of the olive oil controversy: nobody eats 100g of olive oil, but the algorithm scores as though portion size were constant across product types. Defenders argue that per-100g is the only comparable basis available; critics argue it systematically misrepresents products consumed in small quantities.

An updated version of the algorithm has been developed to address several of these criticisms, including harsher treatment of sugary drinks and adjustments for red meat and for fats and oils. Adoption of the revision varies between the countries using the system, which means two packets in different EU states may carry letters computed by different versions.

Design elegance is irrelevant if shoppers ignore the label. The evidence base here is uneven but not empty, and it points fairly clearly in one direction.

The strongest real-world evidence comes from Chile. Its mandatory black warning octagons were introduced alongside restrictions on marketing to children and on child-directed packaging characters. Researchers documented measurable reductions in purchases of products carrying warnings, and analyses indicate the marketing restrictions contributed alongside the labels. Chile is the most-studied case because the policy was mandatory, abrupt, and applied nationally — which makes before-and-after comparison unusually clean.

Nutri-Score has a substantial experimental literature, much of it conducted in controlled or simulated shopping environments. Results generally show improved ability to rank products by nutritional quality and modest shifts in choice. Real-world purchase effects are harder to isolate, partly because adoption is voluntary and uneven.

Traffic light labelling in the UK has high consumer recognition after years of use. Evidence on purchasing effects is mixed, and the voluntary nature of the scheme complicates evaluation for the same reason it complicates Australia's.

The Health Star Rating's most documented finding is about adoption rather than behaviour: voluntary uptake reached only 39 percent in Australia and 36 percent in New Zealand against a 70 percent target, which is what prompted the February 2026 decision to mandate it.

Two conclusions are reasonably well supported by this body of evidence. Mandatory schemes outperform voluntary ones, largely because voluntary adoption is self-selected and the worst products simply do not participate. And warnings appear to outperform scores, plausibly because a warning requires no interpretation — there is no scale to calibrate against and no offsetting to reason about.

This deserves its own treatment because it is the single most important practical finding in the field, and it generalises beyond food.

A voluntary front-of-pack label is adopted selectively by definition. A manufacturer displays the label where it helps and omits it where it does not. Because omission carries no penalty, a shopper cannot distinguish between "this product scored badly" and "this company does not participate in the scheme".

The result is systematic bias. Every label displayed is, on average, better than every label not displayed. The scheme creates an appearance of transparency that the underlying coverage does not support — and in that specific respect a partially adopted voluntary scheme can be worse than no scheme, because it invites shoppers to trust a signal that is not what it appears to be.

Australia's numbers make this concrete. After more than a decade of voluntary operation, the Health Star Rating appeared on roughly four in ten eligible products. The missing six in ten are not a random sample.

Chile avoided the problem entirely by making warnings mandatory from the start. Every product above the thresholds carries the octagon, so absence genuinely means the product is within limits. That single design choice — mandatory application — is probably more important to Chile's measured success than the octagon shape or the wording.

The policy trajectory across jurisdictions follows from this. Voluntary schemes are introduced, adoption plateaus, monitoring documents the shortfall, and a mandate follows. Australia has now reached the final stage. The United States is earlier in the sequence, with a proposed rule from 2025 and options for a final regulation being prepared during 2026.

Most of the world still has none, and it is worth being clear about what that actually means for a shopper rather than assuming the worst.

The back-of-pack nutrition declaration exists almost everywhere. Codex Alimentarius, the international food standards body, provides reference standards that most national systems adopt or adapt, and mandatory nutrition declaration is now widespread. A shopper in a country without front-of-pack labelling is not flying blind; the information is present, just not summarised and not on the side of the packet facing them.

What is lost without a front-of-pack system is speed and comparability at the moment of choice. Research on shopping behaviour consistently finds that people spend very little time on individual purchase decisions, and that information requiring calculation or comparison is largely ignored under those conditions. A nutrition panel that requires turning the packet over, finding the per-100g column, and mentally comparing it to another packet is not going to be used by most shoppers most of the time.

The practical workaround in a market without front-of-pack labelling is a fixed comparison habit. Pick one number that matters for the category you are buying — sugar per 100g for cereals and yoghurts, sodium per 100g for savoury snacks and sauces, saturated fat per 100g for spreads — and compare only that number across the two or three products you are choosing between. It takes seconds once habitual and captures most of what a front-of-pack label would have told you.

The second workaround is the ingredient list, which is universally mandatory and which no front-of-pack system reports. It answers the processing and additive questions entirely, and those are the questions that nutrient-based labelling was never built to address.

Four developments are worth tracking, each with a concrete status.

The United States is preparing its first front-of-package regulation. A proposed rule issued in early 2025 would display added sugar and sodium information on the front of packages. The FDA's 2026 priorities commit to assimilating and summarising public comments and preparing options for leadership to decide the content of a final regulation. Public comment reportedly showed broad support for front-of-package labelling, with disagreement concentrated on content and design rather than the principle itself.

Australia and New Zealand are mandating the Health Star Rating. Food ministers agreed in February 2026, FSANZ has been asked to prepare a proposal for the Food Standards Code, two rounds of public consultation are planned, and a decision is expected in 2027.

The American ultra-processed food definition, submitted for final review in August 2026, could eventually introduce a processing dimension that no existing front-of-pack system captures. Every scheme described here is nutrient-based. A processing-based element would be genuinely novel.

The United Kingdom's nutrient profiling model is under review, with consultation anticipated on applying an updated version to advertising and promotion restrictions. Because the same model underpins the traffic light thresholds and the advertising rules, changes propagate across both regimes.

The direction across all four is consistent: toward mandatory schemes, broader coverage, and — potentially — measures that capture something beyond the nutrient panel.

Front-of-pack labelling looks technical and is not. Every system described here has generated substantial political conflict, and understanding why explains the pace of change.

The Nutri-Score dispute in southern Europe is the most visible. Italy has objected strongly, arguing that the algorithm penalises staples of the Mediterranean diet — olive oil, cured hams, aged cheeses — that have both strong cultural significance and reasonable evidence behind them as components of a healthy dietary pattern. The objection is partly about the per-100g basis and partly about whether a single algorithm can fairly assess foods eaten in very different quantities and contexts. Italy has promoted an alternative approach based on portion-referenced nutritional information.

This is not merely protectionism, though commercial interest is certainly present. It points to a genuine methodological problem: a label that scores per 100g cannot distinguish between a food eaten by the tablespoon and one eaten by the bowl.

A second recurring fight concerns which nutrients count. Saturated fat penalties affect dairy and coconut-producing economies. Sodium penalties affect traditional preserved foods. Every threshold has a constituency with a commercial stake in where it sits.

A third concerns mandatory application. Industry associations have generally supported voluntary schemes and opposed mandates, on cost and proportionality grounds. Those arguments have real content — repackaging is genuinely expensive, and disproportionately so for small producers — and they also happen to preserve the selective-disclosure advantage described above.

Chile's warning system faced sustained legal and commercial challenge before and after implementation, including disputes over packaging character restrictions. That it survived, and produced measurable effects, is a substantial part of why other countries have looked at the warning model seriously.

The pattern is that the technical design debate and the commercial interest debate are thoroughly entangled, and claims from any side should be read with that in mind.

It is worth ending on the limitation shared by every system described here.

All of them are nutrient-based. They measure energy, sugar, salt, saturated fat, and in some cases fibre, protein and fruit or vegetable content. None measures the degree of industrial processing, the presence of additives, or anything about how a food was manufactured.

That gap is why an artificially sweetened drink scores well on most front-of-pack systems while being ultra-processed by any reasonable definition, and why a simple food with a naturally high fat content can score poorly. The algorithms are measuring exactly what they were designed to measure. They were simply not designed to measure the thing that a growing body of research suggests matters independently.

This is the intersection to watch. The United States is simultaneously preparing a front-of-package regulation and developing a federal definition of ultra-processed food. Those two workstreams currently run in parallel. If they ever converge — if a front-of-pack label carried both a nutrient signal and a processing signal — it would be the first genuinely new idea in food labelling in a decade.

Until then, the front of the packet tells you about nutrients and the back tells you about ingredients, and the two answer different questions. Reading only the front means accepting an answer to a question you may not have been asking.