Walk into any grocery store, and chances are, most of what you see is ultra-processed—foods engineered for convenience but often lacking real nutrition. These products make up nearly three-quarters of what’s on the shelves and have been linked to rising rates of obesity, heart disease, and diabetes.
But what exactly makes a food “ultra-processed”? The answer is murky, leaving shoppers confused and policymakers silent on the sidelines.
In a sign of the times, AI may help provide some answers.
The Challenge of Defining Ultra-Processed Foods
Despite the growing evidence linking ultra-processed foods to chronic diseases, even experts struggle to agree on a clear definition.
For a food to be ultra-processed, it must have been chemically altered, said Albert-László Barabási, co-author of a new study on processed foods and co-founder of Foodome—a project mapping the chemical components of the human diet. Certain orange juices labeled “natural,” for instance, are broken down into three separate components, stored individually, and later recombined, thus making them ultra-processed, he said in a press release.
Barabási said that food packaging doesn’t reveal whether a product is ultra-processed. The USDA tracks only a limited number of nutritional components, and the FDA mandates reporting for just about 12 nutrients, leaving much of what’s in our food hidden from view.
This lack of transparency underscores the importance of clear categorization. To help consumers understand food processing, researchers developed the NOVA classification system. Widely regarded as the gold standard for assessing food processing, this method sorts foods into four categories based on how much they’ve been altered from their natural state. Here’s how it breaks down:
- Unprocessed or Minimally Processed Foods – Whole foods that are fresh, dried, ground, pasteurized, or frozen with no added ingredients, such as fruits, vegetables, meat, eggs, and milk.
- Processed Culinary Ingredients – Substances derived from whole foods used in cooking, such as salt, sugar, oils, and butter.
- Processed Foods – Foods made by combining whole foods with culinary ingredients, such as canned vegetables, fresh bread, cheese, and smoked meats. These typically have added salt, sugar, or fat but still retain most of their original food structure.
- Ultra-Processed Foods – Industrial formulations with little to no whole food, containing additives such as emulsifiers, stabilizers, artificial flavors, and preservatives. Examples include sugary cereals, packaged snacks, frozen meals, soda, and instant noodles.
NOVA’s Impact
While the NOVA classification system has been a game-changer for public health research, it is not without flaws, says computational physicist Giulia Menichetti. It groups foods based on processing levels rather than their actual nutrition. That means a fortified breakfast cereal with added fiber and vitamins gets the same “ultra-processed” label as a sugary snack cake—even though the cereal is clearly a better choice.
These inconsistencies aren’t unique to NOVA. Menichetti notes that most food classification systems rely on human interpretation, which can lead to varying conclusions among experts. Without clear, standardized criteria, it remains difficult to define what truly qualifies as ultra-processed—and even harder to incorporate that definition into public health guidelines.
Despite these concerns, many experts still consider NOVA a good starting point. “If you avoid foods in the ultra-processed category, you’re probably making healthier choices overall,” nutrition researcher Nick Norwitz told The Epoch Times. “But not all ultra-processed foods are created equal. There’s nuance in this category.”
What Is GroceryDB?
Enter GroceryDB, a new tool created by researchers at Northeastern University. By analyzing ingredient lists and nutrition labels from over 50,000 food products, GroceryDB ranks foods based on how processed they are. At a time when ultra-processed foods dominate both diets and debates, GroceryDB may offer shoppers a way to take control of what’s in their cart—and their bodies.
Led by Menichetti and published in Nature Food, the GroceryDB study aims to simplify nutrition’s complexities and help consumers make healthier choices.
At the heart of GroceryDB is the FPro algorithm, a machine learning tool that evaluates how processed a food is. Fresh produce ranks near 0, while ultra-processed items such as frozen dinners or sugary snacks climb closer to 1. To make it easier for shoppers, these scores are converted into a simple 0–100 scale on the TrueFood Dashboard.
“Much of nutrition research still relies on manual curation, but our study highlights how artificial intelligence and data science can scale these efforts up,” Menichetti told The Epoch Times.
Unlike other systems that focus on a single nutrient, such as sugar or fat, FPro analyzes a food’s entire nutrient profile. This holistic approach allows the tool to spot patterns that reveal hidden industrial processing—even in foods marketed as “healthy,” such as fortified cereals.
For example, a simple whole-grain bread with no additives might score 31, signaling minimal processing. Add resistant starches or fibers, and the score jumps to 73. A heavily processed version with corn syrup and preservatives could max out at 99.
GroceryDB relies on NOVA as the foundation for training its machine-learning model, raising questions about whether the tool represents a true shift in food classification or simply builds upon the existing framework in a more automated way.
“Food labels often mislead or overwhelm,” said Menichetti. “GroceryDB translates complex data into clear, actionable insights for consumers.”
The algorithm also evaluates individual ingredients in products. Oils such as olive or flaxseed score low for minimal processing, while palm and soybean oils score high as they are ultra-processed. Even similar-looking products can differ dramatically. One cheesecake with 14 ingredients and five additives might score far lower than another with 43 ingredients and 26 additives.
“Our goal is to help people make sense of the overwhelming choices they face in grocery stores,” Menichetti said. “By showing how processed a product really is and the available alternatives, we’re giving consumers a tool to make healthier, more informed decisions.”
But even FPro has its limits. Norwitz points out that we still don’t fully understand how industrial food processing changes food at a chemical level. “There’s a lot we don’t know,” he said. “The hidden effects of food processing—what we call the ‘dark matter of nutrition’—could be critical to understanding the true impact of ultra-processed foods on health.”
The Cost of Convenience
Ultra-processed foods aren’t just prevalent—they’re cheap. A can of processed soup might cost as little as $1 per serving, while preparing a homemade alternative with fresh ingredients could cost triple that or more. For shoppers on tight budgets, this affordability gap makes ultra-processed foods the easiest choice.
Menichetti’s research highlights just how steep this gap is. Highly processed soups and stews are nearly 70 percent cheaper per calorie than their minimally processed counterparts. Across most food categories, every 10 percent increase in processing trims about 9 percent off the cost per calorie—a cost advantage that’s hard to ignore when feeding a family.
This affordability often comes from the use of cheap, calorie-dense ingredients like refined grains, seed oils, and high-fructose corn syrup. Federal subsidies have significantly lowered the production costs of corn, soybeans, and wheat, leading to their widespread use in processed foods. These ingredients are inexpensive to produce at scale but add little nutritional value despite their caloric load.
For families in marginalized communities—where access to fresh produce or whole foods is limited—ultra-processed options often become the default. A 2022 study in The American Journal of Clinical Nutrition found that people who struggle to afford enough food get a larger share of their daily calories from ultra-processed foods compared to those who have steady access to nutritious meals. This reliance on cheaper, calorie-dense, but nutrient-poor foods creates a cycle of poor nutrition and chronic health issues.
GroceryDB’s creators hope the tool can level the playing field by helping consumers identify less processed alternatives within their price range. By offering a clear breakdown of ingredient contributions and processing scores, GroceryDB empowers shoppers to make informed, cost-conscious decisions.
“GroceryDB currently provides the data to systematically identify cost barriers across different food categories and grocery stores,” said Menichetti. “It enables consumers to discover more affordable options within the same category that have comparable FPro scores.”
By making affordable, minimally processed options more accessible, tools like GroceryDB could help close the gap in public health disparities while encouraging healthier eating habits.
Reshaping the Food Landscape
“Even small changes in dietary habits can have a big impact,” said Menichetti. “If we help consumers nudge their choices toward less processed options, we can begin to address chronic health issues tied to ultra-processed foods.”
But GroceryDB isn’t just about better grocery trips. According to Menichetti, the tool has the potential to reshape public health by transforming how policymakers and the food industry approach nutrition.
“This initiative is not just for consumers—it also speaks to agencies, governments, and grocery stores,” Menichetti said.
Policymakers could use GroceryDB to identify areas where healthier options are scarce or too expensive. Grocery stores might then use it to adjust their layouts, stock more minimally processed items, and encourage healthier choices. These changes could have a meaningful impact, especially in communities with limited access to nutritious food.
With food processing under increasing scrutiny, tools like GroceryDB—alongside policy changes and consumer awareness—could help reshape the way we shop and eat for the better.
“GroceryDB is just the beginning,” said Menichetti. “We’ve barely scratched the surface.”
