There was a period during my concussion recovery when I ate the same lunch nearly every day: a large spinach salad with plenty of nuts and some whole grains. According to the conventional recommendations, this was just about the perfect meal: nutrient-dense, plant-based, and whole foods, exactly what a health-conscious person eats when they are trying to get well. Yet, every single afternoon, I would get the same predictable energy crash accompanied with a feeling of heaviness, and brain fog that would render me out of commission for hours. I spent months eating this way and getting the same reactions, not drawing the obvious conclusion that the salad could be causing these symptoms. All the experts were seemingly in complete agreement that this was just about the best thing I could be doing for my body so why question it.
I think about that period often now, because friends and family regularly pull out their phones to show me an app called Yuka. This app lets you scan the barcode of food items at the grocery store and it gives you a health score between 0-100. When I scan the ingredients in that same spinach salad that caused me so many issues, the app gives extremely high scores, perpetuating this idea that this is an extremely healthy choice. Compare that with the foods that actually supported my healing like grass-fed ground beef, and Yuka gives it a far lower score. How could this app’s ratings be so incongruent with my own healing experience? This question prompted a deep dive into how Yuka actually forms its recommendations, what it does well, and where it goes wrong.
What Yuka Does Well
I want to be fair here, because the instinct behind Yuka is a good one. Most people want to be healthier and almost nobody has the time or training to keep up with all the research. The modern grocery store is engineered to make evaluation difficult. An app that scans a barcode and gives you a straight answer helps you see beyond common marketing ploys and understand exactly what’s in your food.
When evaluating common food additives, it genuinely has a practical use case. Two products can look identical on the shelf while one contains compounds worth avoiding. If you saw carrageenan on one label and ascorbic acid on another, you might reasonably assume both are industrial chemicals to be avoided. In fact ascorbic acid is vitamin C, used as a preservative, and carrageenan is a compound with real evidence for damaging the gut lining. Yuka distinguishes between them and links you to why. That is a useful service, and it is the part of the app I would keep.
The trouble is everything else. Because when Yuka moves from “what has been added to this food” to “is this food good for you”, it starts confidently telling people that nutrient-dense foods are problematic, often recommending anti-nutrient rich foods, highly processed foods, and refined starches instead.
A Rating is an Opinion
Here is the thing most users never think about: a score out of 100 looks like a measurement, but it is actually the output of a formula. The formula they use encodes a very specific set of beliefs about what makes food healthy.
Yuka’s score is 60% nutritional quality, 30% additives, and a 10% automatic bonus for organic certification. That 60% runs on Nutri-Score, a system developed by French public health researchers. Once you see what Nutri-Score actually measures, you can begin to understand how nutrient dense, ancestrally consistent, whole foods could be demonized in replacement of refined starch.
The entire rating system has categories that add points to the score of a given food, and categories that take away points for a given food. A given food receives points for its fibre content, protein content and for the percentage of its content consisting of fruit, vegetables, legumes and nuts. A food loses points for having excess “energy”, sugars, saturated fat, and salt. Add points from the first list, subtract points for the second list, and you get the nutritional quality score.
Read that list again and notice what is missing.
There is no term for micronutrient density, nutrient bioavailability (whether your body can actually absorb what’s in the food), none for anti-nutrients (compounds that block nutrient absorption and damage tissue), and none for fat stability under heat. Instead it uses these oversimplified six variables to determine whether a food is healthy, and four of them are things the model wants you to eat less of.
Now let’s pay attention to what is actually included in this model. Saturated fat and salt, two things humans consumed throughout our evolution and which form part of any well-formulated ancestral diet, are scored as direct penalties. Excess sugars are penalized, correctly, but no other refined carbohydrate is, meaning refined starch with an essentially identical blood sugar impact passes through entirely unnoticed.
Looking closer at how fat in particular is handled, you’ll see where the formula does arguably the most damage. Nutri-Score has exactly one term for fat, and it’s saturated fat, counted in absolute grams. There is no variable for oxidative stability, nothing that asks whether a fat holds up under heat or breaks down into the toxic aldehydes that do the actual cellular damage. In fact, the model actively demonizes the nutrient-dense fats that are most heat stable. So the model penalizes butter, cream, and tallow, fats humans have eaten for millennia, while modern industrial seed oils rich in these toxic aldehydes pass through untouched.
Then there is the bonus category where plant foods get an arbitrary boost. A food earns credit for being composed of fruit, vegetables, legumes and nuts, regardless of its actual biochemical profile or anti-nutrient load. There is no equivalent category that meat, eggs, or organ meats can qualify for. Humans evolved eating a substantial amount of meat and associated fat, and these provide the most robust bioavailable nutrition for the human body. Yet this rating system arbitrarily decides that the less bioavailable and ancestrally consistent nutrition from plant-based foods gets a boost.
I want to be clear that I don’t think anyone involved acted in bad faith. Nutri-Score was built by public health researchers with genuine intentions, and it faithfully encodes the mainstream nutritional consensus. That’s precisely the problem. That consensus is built on nutritional epidemiology and its food frequency questionnaires, weak associations, and statistical adjustment, ignoring the hard sciences of what actually drives metabolic health. I’ve unpacked that at length in my previous post: Why Nutrition Research is Unreliable. Yuka isn’t corrupt, It’s orthodox. And when you encode a flawed orthodoxy into an algorithm and hand it to sixty million people, the results are entirely predictable.
What This Looks Like in Practice
I took the app to my local grocery store to test exactly what this algorithm looked like in practice. Here are some of the most egregious results:
White pasta: 90
Orange Juice: 53
The juice is flagged as too caloric and too sweet, but white pasta converts to glucose just as fast, producing the same insulin response, the same glycation damage, the same downstream inflammation, only without any of the antioxidants or micronutrients found in orange juice. The absence of additives does not make it a healthy choice, nor does the presence of 5g of fibre in an over 60g serving of refined starch. Orange juice can certainly be a concentrated source of sugar and can be problematic in excess, but white pasta really isn’t any different. This single comparison exposes the app’s blindness to one of the most damaging components of the modern diet.
Almond Milk: 88
Grass-fed whole sheep’s milk: 78
One is a nutrient-dense whole food humans have consumed since the dawn of animal husbandry. The other is a processed nut product carrying oxalates and phytates that bind minerals and, in excess, contribute to kidney stone formation and joint pain.
Heavy cream: 45
A product with no additives at all, downgraded purely for its saturated fat. Its suggested replacement? Soy cream, built on unstable polyunsaturated fats that oxidize readily under heat producing harmful toxic aldehyde byproducts.
Organic Baby Spinach: 100
Organic Grass-Fed Ground Beef: 49
Meat provides every essential amino acid and fatty acid a human requires, along with the most bioavailable forms of vitamins and mineral that support every body system. Spinach delivers an oxalate load that, consumed daily by a sensitive person, contributes to joint pain, kidney stones, and impaired mineral absorption.
This one was not a hypothetical for someone like me. A 100 out of 100 does not say “this is fine occasionally”, it says “eat more of this”. That is exactly what I did, and its this same advice that has people prioritizing the complete wrong foods in their diets.
Dill pickle chips: 32
Pork Rinds: 30
Dill pickle chips are correctly rated poorly, but for entirely the wrong reasons. It is listed as too caloric, and containing too much sodium. Not a word about the vegetable oil listed second on the label, which were likely sitting in a commercial fryer and accumulating toxic aldehydes. Apply that same logic and you understand why pork rinds, a whole food consisting of pork skin, lard, and salt, scores even worse.
The Cost of Getting This Wrong
Followed faithfully over months, Yuka steers you toward anti-nutrient rich plants, processed plant products, and refined starches, and away from the nutrient-dense animal foods that formed the backbone of human nutrition for the overwhelming majority of our evolutionary history. It builds a mental model in which plants are inherently virtuous and animal foods are something to be limited.
It doesn’t just improperly rate individual foods. It reshapes a shopping cart, and then a diet, and then a body.
The Importance of Context
Setting aside the questionable formula behind their calculation, the idea that all foods can be given a healthiness score deserves serious scrutiny. For all plant-based foods, healthiness isn’t an intrinsic property, but a relationship between that given food, where it was grown, a person’s microbiome, and their unique life circumstances.
Seasonality alone breaks the model entirely. Fruit grown locally in August is a very different proposition than one shipped across several borders in the middle of February. Dietary context plays an even more important role in determining the relative value of certain foods. An athlete struggling to eat enough or someone with a sluggish thyroid might require some more palatable carbohydrates to help stimulate appetite and eat enough food to sustain their metabolic requirements. Conversely, someone with blood sugar dysregulation, weight gain, or inflammation would find those same foods working against their goals. Having to give one food the same score for these two individuals is an impossible task, and is partially why they default to such an oversimplified scoring system.
What to Do Instead
You don’t need a better app. You need to stop outsourcing the judgment.
These tools exist to help you navigate the packaged aisles and the honest answer is that you should mostly not be in them. Shop the meat counter and the produce section, and the problem largely dissolves.
For anything that does come in a package, four rules will outperform any scanner:
- If the ingredient list runs longer than about five items, put it back.
- If you can’t recognize an ingredient or picture the whole food it came from, put it back.
- If it contains industrial seed oils, put it back.
- If the carbohydrate load doesn’t fit your goals, put it back.
And for everything else, ask the only question that has ever actually mattered: “is this consistent with what I evolved to eat?”
The foods worth building a diet around often don’t have barcodes at all and are easily recognizable at first glance. Meat, eggs, fish, butter, and seasonal produce should form the vast majority of what you eat. I cover the science supporting this in my post: The Ancestral Dietary Framework.
Just as important is asking yourself whether the foods you’re eating are really making you feel your best. If I had focused on my own body intuition rather than listened to pseudoscientific mainstream guidelines, it wouldn’t have taken me months to stop eating all that spinach. Don’t let apps like Yuka tell you what is good for you. Understand what your biology evolved to eat, experiment within that framework, and you’ll feel much better than when you were letting an app tell you what’s good for you.