Processed food scanner evidence is a classification problem, not a shopping trophy. Apps compress NOVA, additives and house rules into one glance. Experts do not always agree on the same yogurt. This is a source-reviewed guide, not a hands-on lab test.
A red “ultra-processed” badge at the fridge door feels like a lab result. It is a label read through a rulebook. Processed food scanner evidence depends on whether humans can apply that rulebook the same way twice.
This page sits beside the best meal planning apps hub. For claimed product features see the Ivy review and the Yuka review. Photo calorie tools such as Foodvisor answer a different question. We do not invent a scanner league table.
Do not treat a processing class as a diagnosis. Fortified foods, medical nutrition and affordable staples can sit in NOVA 4 and still have a role. If labels fuel fear, step back.
What NOVA is designed to do
Monteiro and colleagues (Public Health Nutrition, 2019) describe four groups. Group 1 is unprocessed or minimally processed food. Group 2 is culinary ingredients such as oil, sugar and salt. Group 3 is processed foods that combine those. Group 4 is ultra-processed industrial formulations, often with little intact group-1 food left.
NOVA does not score protein quality, fortification or price. A whole-grain extruded cereal and a confectionery bar can share group 4. That is a feature of the system. It is also the main consumer confusion.
SACN’s July 2023 statement, and its April 2025 rapid update, reviewed eight processing classifications. Only NOVA met the committee’s initial screening criteria for possible UK use. SACN still worried about practical application. Some NOVA assignments clash with nutrient-based ideas of a “healthier” food. The 2025 update said higher (ultra) processed intake still looks concerning. It also said the evidence cannot yet separate processing itself from the fact that many of these foods are high in energy, saturated fat, salt or free sugars.
What processed food scanner evidence actually shows
Reliability is the first evidence question. If trained people disagree, a phone will disagree too.
Braesco’s team asked French food and nutrition specialists to assign NOVA groups. One list had 120 marketed products with ingredients (159 evaluators). One list had 111 generic foods without ingredients (177 evaluators). Fleiss’ κ was 0.32 and 0.34. That is poor agreement. A cluster of marketed foods was mostly called NOVA 4. Other clusters were mixed. The authors said current criteria do not allow robust, functional assignment.
Bleiweiss-Sande and colleagues (Nutrients, 2019) coded the top 100 foods eaten by US children in NHANES 2013–2014. Inter-rater rank correlation was 0.97 for the UNC system, 0.78 for IFIC and 0.76 for NOVA. Sodium and added sugars helped predict “highly processed” classes. The authors said the systems may not flag nutrient-dense processed foods children actually eat.
Trained double-coding can look better. A 2023 American Journal of Clinical Nutrition analysis of preschooler recalls reported Cohen’s κ 0.75 after certified coder pairs used NDSR. That is a research workflow, not a shopper tapping a barcode.
Hall and colleagues (Cell Metabolism, 2019; NIH news release) ran an inpatient crossover in 20 adults. An ultra-processed menu and a minimally processed menu were matched for presented nutrients. People ate more energy on the ultra-processed menu and gained weight in two weeks. That trial supports a mechanism story. It does not validate a consumer scanner.
We found no peer-reviewed paper that takes Ivy’s house rating or Yuka’s 100-point mark and compares them with expert NOVA coding on a stated product set. Ivy’s terms (captured 13 September 2026) say the app uses Open Food Facts and manufacturer labels, and that the publisher does not analyse composition in a lab. Yuka’s help page is a Nutri-Score-plus-additives-plus-organic mix, not a NOVA engine.
| Question or claim | Evidence source | Study type | Population | Reference standard | Outcome | Key finding | Limitation | Applicability |
|---|---|---|---|---|---|---|---|---|
| Can specialists agree on NOVA? | Braesco 2022, EJCN | Expert assignment survey | 159–177 French specialists; 231 foods | NOVA criteria as published | Fleiss’ κ | κ = 0.32 (marketed) and 0.34 (generic) | One country; one survey era | A barcode class can be an opinion |
| Is NOVA as repeatable as other systems? | Bleiweiss-Sande 2019, Nutrients | Dual coding | Top 100 child foods, NHANES | NOVA, IFIC, UNC | Inter-rater ρ | NOVA ρ = 0.76; UNC 0.97; IFIC 0.78 | Child foods only | Edge cases will split scanners |
| Is NOVA suitable for UK policy? | SACN 2023 statement; 2025 update | Evidence review | Published classifications and health studies | SACN screening criteria | Policy suitability | Only NOVA passed the screen; application concerns remain; processing vs nutrients still unclear | Observational dominance | Use as a research lens, not a legal stamp |
| Do UPF menus raise short-term intake? | Hall 2019, Cell Metabolism / NIH | Inpatient crossover | 20 adults | Weighed ad libitum menus | Energy and weight | Higher energy intake and weight gain on the UPF menu | Small, short, ward setting | Supports caution on UPF patterns, not on one scan |
| Does a nutrition score equal NOVA? | Romero Ferreiro 2021, Nutrients | Cross-classification | Spanish products | Nutri-Score and NOVA | Class overlap | UPF appeared in every Nutri-Score class, including A | Older algorithm | A green pack score can still be NOVA 4 |
Additives are not a personal risk colour
Many scanners paint E-numbers. Yuka assigns green-to-red risk and can cap a product at 49/100 if it marks an additive high-risk. That is a house reading of EFSA, IARC and selected papers. It is not a personal exposure assessment.
An approved additive has a permitted use. An IARC hazard classification is not the same as a dietary risk at the dose in one yogurt. Observational links from one cohort should not become a scarlet letter on every emulsifier.
Ivy’s public copy lists seed oils, pesticides, microplastics and heavy metals among score inputs. Those are different scientific questions. A barcode cannot measure microplastics in the pack you hold unless a lab result is attached. If the method page does not show a lab, treat the extra marks as inferred, not assayed.
What a scanner cannot see
Your portion. A class is per product, not per gram you eat.
Your week. One NOVA 4 bread does not describe a pattern. SACN’s caution is about habitual diets.
Home recipes. A photo of leftovers is not a barcode. Processing class from an image is still experimental. We do not treat it as settled.
Nutrient needs. Protein after illness, iodine in pregnancy, or a low-potassium clinical diet can require products a scanner paints red. Our everyday protein intake guide stays in food language for that reason.
How to use a scanner without a purity test
Use it to notice formulation. Long additive lists and sweetened drinks are fair flags. Then read the Eatwell plate for the week. Swap one daily sugary drink. Add a bag of frozen vegetables beside a ready meal. That is a pattern change. It is not a ban.
Compare two similar products. Two baked beans. Two yogurts. If both are NOVA 4, the lower sugar and salt pack still matters. Nutrient profile and processing are complementary, as Sarda, Julia and colleagues have argued for Nutri-Score updates.
Ignore health promises on the store listing. Ivy’s Play copy (captured 13 September 2026) claims users lose weight and fix anxiety. Those are marketing sentences. They are not trial results. We omit them as facts.
What remains unverified
We cannot verify 2026 Ivy or Yuka accuracy against a gold-standard NOVA panel. We cannot say a red scan causes disease. We cannot convert a week of green scans into a health outcome.
Processed food scanner evidence is strongest at the level of systems (NOVA reliability, SACN caution, nutrient-score mismatch). It is weakest at the level of one branded app. Keep those layers separate.
Frequently asked questions
Is every ultra-processed food unhealthy?
Why do two scanners disagree?
Does a good Nutri-Score mean the food is not ultra-processed?
Can I trust additive risk dots?
Should I drop every NOVA 4 product?
Did Vitality Ledger scan a supermarket aisle?
Sources
- 1. Ultra-processed foods: what they are and how to identify them
- 2. SACN statement on processed foods and health
- 3. Ultra-processed foods: how functional is the NOVA system?
- 4. Robustness of food processing classification systems
- 5. Study finds ultra-processed diets cause excess calorie intake and weight gain
- 6. Two dimensions of nutritional value: Nutri-Score and NOVA
Guidance changes. Figures were checked against the sources above at the time of review; always confirm current advice with your GP, pharmacist or clinician.
Image credits
- Photo: Photo by Klaus Nielsen on Pexels / Openverse
- Photo: Photo by Jud McCranie on Wikimedia Commons / Openverse
- Photo: Photo by Daniel Schwen on Wikimedia Commons / Openverse
- Photo: Photo by Olu Eletu on Unsplash via Wikimedia Commons / Openverse
- Photo: Photo by Shixart1985 on Wikimedia Commons / Openverse
Was this useful?
Anyone can react — no account needed.
Discussion
0 comments · Name and email only · Email is never shown