Is Nutri-Score really reliable?

1. Reliability depends on the question asked

Nutri-Score has an operational purpose: summarize a nutritional profile per 100 g or 100 mL and make rapid comparisons easier. Its formula is public and reproducible when the necessary data are available. Reproducibility does not mean the result captures every property of a food.

Four levels of evidence must then be separated. Comprehension asks whether consumers rank products more accurately. Behavior examines what they choose or buy. Profile validity looks for associations between average dietary quality and health. Demonstrating the logo’s impact would ultimately require showing that exposure produces lasting changes in diet and then in health.

An online experiment can answer the first question without demonstrating the fourth. A cohort study can support the profile’s relevance without testing the package logo. Reliability is therefore not an unconditional yes or no; it depends on the intended use and study design.

2. Evidence is strongest for immediate comprehension

An online experiment involving 12,391 adults in 12 countries compared five front-of-pack labels. After randomization, Nutri-Score produced the largest improvement in participants’ ability to rank three products correctly by nutritional profile, ahead of the other formats tested. Effects on reported choices were smaller and varied more across countries and food categories. Egnell et al., 2020 — randomized experiment in 12 countries

The finding directly supports the simplification function: a grade and color help people rank options. It does not show that participants bought those products, sustained the choice over time, or improved their health.

Participants were recruited online and saw only three food categories. The average effect therefore cannot guarantee identical comprehension for every person, category, or busy store environment. It nevertheless matches the logo’s most immediate objective.

3. In stores, the effect exists but is much smaller

A French trial by Dubois and colleagues analyzed 1,668,301 purchases involving 1,266 products across four in-store categories. Nutri-Score increased purchases from the most nutritious third of each category by approximately 14%, with no significant effect on the other two thirds. The nutritional quality of the basket of labeled products improved by approximately 2.5%. Dubois et al., 2021 — real-world trial of supermarket shopping

The authors estimated effects approximately 17 times smaller than in their laboratory setting. That gap does not erase the result; it shows how price, habits, availability, promotions, and preferences dilute a logo’s influence in real life.

The study covered four product categories, selected stores, and a specific period. It measured neither participants’ entire diets nor disease outcomes. The appropriately scaled conclusion is a modest shift in some purchases, not an automatic transformation of shopping baskets.

4. Evidence syntheses find an effect on choice, but many studies are artificial

A network meta-analysis of 135 studies estimated that Nutri-Score increased the odds of choosing or purchasing a healthier option, with an odds ratio of 1.51 and a 95% confidence interval of 1.08–2.11. Other warning-label formats performed better at reducing certain less healthy purchases. Song et al., 2021 — network meta-analysis of 135 studies

Approximately 95% of the studies were conducted in laboratory settings; only eight were field studies, more than half had a high risk of bias, and none tracked a long-term outcome. The average estimate should not become a precise prediction for purchases in France in 2026.

A 2025 review focused on Nutri-Score likewise judged actual changes in habits and health to be small or nonsignificant in many studies, with modest improvements in shopping baskets, particularly for sodium and saturated fat. The review is debated and does not settle the issue; above all, it reinforces the need to separate comprehension, intention, and sustained behavior. 2025 review — state of the evidence specific to Nutri-Score

5. Cohort studies support the profile, not a causal effect of the logo

Cohort studies reconstructed a dietary index from the underlying nutritional model and found that a less favorable average profile was associated with higher disease or mortality risk. A 2023 meta-analysis combined these relationships and supports the model’s usefulness for characterizing overall nutritional quality. 2023 meta-analysis — validity of the nutritional-profile model and health outcomes

These analyses do not randomize exposure to package labels or compare people who see the logo with those who do not for years. They assess average dietary composition and retain the usual limitations of observational data: self-reported measurements, residual confounding, and changes in habits.

They therefore do not prove that buying one A-rated product prevents disease or that seeing the logo reduces mortality. They support a different conclusion: the nutrients included in the profile capture one dimension of diet that is relevant to health when studied across the overall diet.

6. The main blind spot is confusing a profile with a product or an entire diet

Belgian and Swiss authorities note that Nutri-Score does not account for the serving actually consumed and does not evaluate every dimension of a food. The grade does not summarize additives, contaminants, allergens, processing level, sustainability, many micronutrients, or individual needs. FPS Public Health Belgium — uses and limitations of Nutri-Score, Swiss Federal Office — scope and limitations of Nutri-Score

An A- or B-rated product may be ultra-processed, easy to overconsume, or poorly suited to a particular meal. A D- or E-rated food may be used in a small amount, serve a culinary purpose, or fit within a varied diet. Those examples do not make the calculation fraudulent; they show that it measures only one dimension.

The most useful comparison is between products that can substitute for each other. Using the grade to choose between two breads is more informative than ranking bread, oil, and cheese as though they served the same purpose. Frequency, serving size, and the rest of the diet then complete the decision.

• Read Nutri-Score and Nova as two complementary dimensions →

7. The algorithm is public, but its conventions remain debatable

The official calculation defines its nutrients, thresholds, and exceptions. This transparency makes the result verifiable; it does not turn every modeling decision into a natural law. A 100 g basis supports comparability but ignores serving size, while one letter hides numerical differences within the same grade.

The 2022 revision report also acknowledges a data constraint: added sugar and the proportion of whole grains are not declared uniformly or mandatorily on European labels. They therefore cannot be used directly across the general calculation even though they could help distinguish certain products. Scientific committee, 2022 — available data and algorithm limitations

The update that took effect in 2025 tightened thresholds for sugar, salt, and fiber, revised beverage scoring, and better differentiated fats. It addresses several known limitations without eliminating the tradeoffs inherent in a summary logo.

• Understand the changes in the updated algorithm →
• Review the official formula and thresholds →

8. The word “scandal” conceals very different controversies

A study by Mialon and Mialon identified 367 public examples of political activity by five food-industry stakeholders between March and August 2015, including information dissemination, coalition building, and interventions concerning evidence. It documents strategies used during the French debate; it does not demonstrate that one action caused the system’s adoption, determined the algorithm’s form, or delayed a European decision. Mialon and Mialon, 2018 — documented political activities around the French debate

Several favorable studies include the logo’s designers or advocates as authors, while several reviews disclose food-industry ties. Those facts require close examination of study protocols, data, and conflict-of-interest statements. They are not sufficient by themselves to declare a study false or fraudulent.

In 2026, Oqali data showed a slight decline in the share of sales volumes covered, particularly among some national brands, despite an increase in the total number of participating companies. The data do not identify each company’s reason for withdrawing and cannot establish that every departing brand had unfavorable grades. INRAE and Oqali, 2026 — changes in brand participation

None of these observations justifies the shortcut “Nutri-Score is a lie.” They support independent monitoring, public data, algorithm updates, and clear explanations of what each study actually measures.

9. ANSES’s 2017 opinion is a historical snapshot, not a 2026 verdict

On January 31, 2017, ANSES concluded that the five nutrition-information systems it examined lacked sufficient evidence to demonstrate their relevance for reducing disease incidence. The agency did not declare Nutri-Score dangerous or prove that it was ineffective for every other purpose. ANSES, 2017 — evidence level for nutrition-information systems

At that time, France had not yet adopted the logo and the available evidence was more limited. Since then, comprehension experiments, purchasing trials, and cohort studies have expanded the evidence base. They support certain stages more strongly but still do not provide a long-term trial showing that the logo itself prevents disease.

The accurate reading holds both points together: the older opinion cannot serve as a definitive condemnation, and the accumulation of intermediate evidence should not be described as direct clinical proof.

10. The Lactalis case did not cancel Nutri-Score

Groupe Lactalis sought annulment of the March 14, 2025 decree and the provisions classifying milk, drinkable yogurt, and flavored or chocolate milk drinks as beverages. On June 16, 2026, France’s Council of State rejected several arguments, including those concerning consultation, misleading presentation, objectivity, and discrimination. Council of State, decision no. 508230 of June 16, 2026

The Council of State nevertheless found that two serious questions of European-law interpretation should be referred to the Court of Justice of the European Union. They concern, in particular, whether Article 35 of the Food Information to Consumers Regulation permits a letter-and-color summary and the inclusion of elements that do not all appear in the mandatory nutrition declaration. The Council of State therefore stayed the proceedings.

As of July 26, 2026, the CJEU had not issued its judgment, and the French decree remained in force. It is factually inaccurate to say that Lactalis won, that the Council of State definitively upheld every rule, or that it declared Nutri-Score illegal. This is an unresolved legal question, not a scientific experiment on the logo’s effects.

11. Practical verdict: reliable as a filter, insufficient as a judge

Use Nutri-Score first to compare products that serve the same purpose. When grades differ, check the nutrition facts to see whether the difference comes from salt, sugar, saturated fat, fiber, or another criterion. Then consider the ingredient list, processing level, serving size, and frequency.

For sports nutrition, the grade does not know your energy needs, protein target, meal timing, or digestive tolerance. It can improve one choice, but the full day determines whether intake is coherent. Dietary guidelines and individualized support, when needed, remain more important than the logo.

• Comprehension: convincing evidence for quickly ranking the products tested.
• Purchases: a favorable but modest and variable signal in real-world settings.
• Health: useful associations with the dietary profile, without causal proof that logo exposure improves health.
• Scope: no overall assessment of processing, additives, or serving size.
• Law: European questions remain pending, with no final ruling as of July 26, 2026.
• Use: compare, verify, and contextualize—never prohibit a food based on one grade.