For packaged foods, the protein number is printed right on the nutrition label, calculated per serving size. For unpackaged foods like fresh produce, bulk grains, or a home-cooked meal, you need a food composition database or a tracking app to estimate protein content by weight. Both approaches ultimately trace back to the same laboratory chemistry: measuring nitrogen and multiplying by a conversion factor. That simple multiplication, though, hides a surprising amount of imprecision, and understanding where the numbers come from changes how much you should trust them.
What the Nutrition Label Actually Tells You
The protein figure on a food label is not measured directly as protein. In most regulatory systems, manufacturers determine protein by measuring total nitrogen content in a food sample and multiplying it by a standard conversion factor, typically 6.25. This method dates back to the 1880s and rests on the assumption that protein contains about 16 percent nitrogen on average, so dividing 100 by 16 gives you 6.25. It is a rough estimate dressed up as a precise number.
The problem is that not all nitrogen in food comes from protein. Foods contain non-protein nitrogen from compounds like free amino acids, nucleotides, creatine, and urea. In milk, for instance, non-protein nitrogen accounts for anywhere from 3 to 25 percent of total nitrogen depending on the source and processing conditions.1International Dairy Journal. Non-protein nitrogen in milk and infant formula and its relevance in infant nutrition That means the crude protein value on a milk carton includes nitrogen from substances your body does not use as protein. The same issue applies to meat, fish, and especially plant foods.
A study comparing crude protein values to actual amino acid content across food categories found consistent overestimation. The nitrogen-times-6.25 calculation overstated true protein by about 5.5 percent in dairy products, about 15 percent in meat, roughly 21 percent in fish, and around 12 to 14 percent in cereals and vegetables.2Food Chemistry. Determination of protein in foods: comparison of net protein and crude protein (N × 6.25) values So when a label says a serving of canned tuna has 20 grams of protein, the actual amino acid content could be closer to 16 grams. This does not mean labels are useless, but the numbers are better understood as rough guides than precise measurements.
Why 6.25 Is Still the Default
Scientists have known for decades that 6.25 overestimates protein for most foods. In 1931, researcher D. B. Jones published food-specific conversion factors that accounted for differences in amino acid composition across food types. Soy, for example, was assigned a factor of 5.71 because its main storage protein, glycinin, contains more nitrogen per unit of protein than the average.3Journal of the American Oil Chemists’ Society. Calculation of Nitrogen‐to‐Protein Conversion Factors: A Review with a Focus on Soy Protein Milk was given 6.38. Wheat and rice got 5.83. Despite the availability of these more accurate factors, many food labels and databases still default to 6.25 because no universally accepted revision has been implemented across regulatory systems.4PubMed. Toward harmonizing protein data in food composition databases: evaluating perspectives, methods and implications
The situation is even more dramatic for leafy plant foods. A study of 90 plant species found that appropriate conversion factors ranged from about 3.3 to 5.2 when calculated from actual amino acid recovery, with an average around 4.4.5Food Chemistry. Leaf protein contents and nitrogen-to-protein conversion factors for 90 plant species Using 6.25 on spinach or kale, in other words, could overstate their protein by 30 to 40 percent. For most people tracking macros casually, this error is buried inside a minor food category. But for anyone relying heavily on plant-based protein sources, it is worth knowing that the label may paint an optimistic picture.
Estimating Protein in Unlabeled Foods
Fresh fruits, vegetables, meat from a butcher counter, eggs, and anything you cook from scratch will not come with a nutrition label. The most practical tool for estimating protein in these foods is a food composition database. In the United States, the USDA’s FoodData Central is the standard reference. It integrates several datasets, including Foundation Foods with detailed nutrient analysis of common whole foods, the older Standard Reference database, and a branded-food database with label data for over 350,000 products.6PubMed Central. FoodData Central, USDA’s Updated Approach to Food Composition Data Systems Most other countries maintain equivalent databases: the UK has McCance and Widdowson’s, Australia has NUTTAB, and Japan has the Standard Tables of Food Composition.
To estimate protein in a home-cooked meal, you weigh individual ingredients raw, look up the protein per 100 grams in the database, and multiply by the amount you used. A 150-gram raw chicken breast listed at 23 grams of protein per 100 grams would contribute about 34.5 grams. Do the same for each ingredient, add them up, and divide by the number of servings. The math is simple. The hard part is accuracy in weighing and accounting for water loss during cooking.
How Cooking Changes the Numbers
Heat does not destroy protein in meaningful amounts during normal cooking. What cooking does is drive off water, which concentrates the protein per unit of weight. A raw chicken breast that weighs 150 grams might weigh 120 grams after grilling, and the protein per gram of cooked meat is higher simply because you lost moisture. If you weighed the chicken raw and used the raw database value, your estimate is fine. If you weigh it cooked and still use the raw protein-per-gram figure, you will undercount.
This concentration effect applies to all cooked foods. Research on both beef and plant-based burgers has confirmed that protein concentration increases during cooking as water evaporates.7Journal of Agriculture and Food Research. Impact of cooking method on properties of beef and plant-based burgers: Appearance, texture, thermal properties, and shrinkage Food composition databases handle this by listing separate entries for raw and cooked versions of common foods, which is why it matters to pick the right entry for how you weighed the ingredient.
Tracking Apps and Their Accuracy
Most people estimating daily protein intake use a smartphone app like MyFitnessPal, Cronometer, Lose It!, or similar tools. These apps pull from food composition databases and user-submitted entries, and their convenience is unmatched. But how accurate are they?
A systematic review and meta-analysis of validation studies on dietary record apps found that they consistently underestimated protein intake, with a pooled underestimate of about 12 grams per day compared to research-grade dietary analysis software.8Advances in Nutrition. A Systematic Review and Meta-Analysis of Validation Studies Performed on Dietary Record Apps Another study comparing five popular nutrition tracking apps against a research standard found that macronutrient correlations were generally high (0.73 to 0.96 for protein), but most apps produced values significantly lower than the reference.9PubMed Central. Assessment of the accuracy of nutrient calculations of five popular nutrition tracking applications One evaluation found that the app Cronometer was the exception, producing estimates that did not differ significantly from the reference software across macronutrients.10PubMed Central. Mobile Apps for Dietary and Food Timing Assessment: Evaluation for Use in Clinical Research
The underestimation likely comes from a combination of factors: incomplete food databases, user errors in portion estimation, and default serving sizes that do not match what people actually eat. If you are tracking protein closely for athletic performance or medical reasons, it helps to know that your app probably lowballs the number slightly. Using a kitchen scale and selecting database entries carefully will narrow the gap.
How Laboratories Actually Measure Protein
Two classic laboratory methods dominate protein measurement in the food industry, and both work by measuring nitrogen rather than protein directly. The Kjeldahl method, developed in 1883, digests a food sample in sulfuric acid to convert all nitrogen to ammonium, which is then measured. The Dumas method combusts the sample at high temperature and measures the nitrogen gas released. Both give you total nitrogen, which is then multiplied by a conversion factor to yield crude protein.11Europe PMC. Measuring Protein Content in Food: An Overview of Methods
The Dumas method tends to produce results about 1.4 percent higher than the Kjeldahl method on average, though the gap varies by food type.12Analyst. A comparison of the Kjeldahl and Dumas methods for the determination of protein in foods, using data from a proficiency testing scheme Both methods have the same fundamental limitation: they measure nitrogen, not protein. Any nitrogen-containing compound in the food, whether it comes from amino acids, DNA, caffeine, or chitin, gets counted.
The gold standard for true protein measurement is total amino acid analysis, which breaks a food sample down into its individual amino acids and adds them up. This is far more expensive and time-consuming, which is why it is reserved for research rather than routine label verification. A study on beef cuts found that the correlation between total amino acids and crude protein (nitrogen times 6.25) was surprisingly poor, with amino acid totals averaging only about 91 percent of the crude protein value.13PubMed. Total nitrogen vs. amino-acid profile as indicator of protein content of beef That nine-percent gap is entirely non-protein nitrogen being counted as protein.
Near-Infrared Spectroscopy and Rapid Testing
The food industry increasingly uses near-infrared spectroscopy as a fast, non-destructive way to estimate protein content. NIR works by shining light at specific wavelengths through or off a food sample and analyzing how the sample absorbs or reflects that light. Different chemical bonds absorb different wavelengths, and mathematical models trained on reference samples can predict protein content in seconds without grinding, digesting, or destroying the food.14PubMed Central. The Ability of Near Infrared (NIR) Spectroscopy to Predict Functional Properties in Foods: Challenges and Opportunities
NIR is already standard for grading wheat, where protein content determines baking quality and price. One study measuring protein in wheat found that NIR hyperspectral imaging achieved prediction accuracy on par with conventional benchtop NIR instruments, with error margins as low as 0.15 percent.15Journal of Near Infrared Spectroscopy. Quantification of protein in wheat using near infrared hyperspectral imaging: Performance comparison with conventional near infrared spectroscopy The technology is also being explored for plant-based protein products, where mixtures of pea protein, oils, and starches make traditional methods cumbersome.16Journal of Food Composition and Analysis. Near-infrared spectroscopy-based quantification of sunflower oil and pea protein isolate in dense mixtures for novel plant-based products Handheld NIR devices marketed to consumers exist, but their accuracy for mixed meals is still limited compared to laboratory instruments.
Protein Quality Is Not the Same as Protein Quantity
Even if you nail the protein content of a food down to the gram, that number does not tell you how much of that protein your body can actually use. Two foods with identical protein content on a label can deliver very different amounts of usable amino acids. This is where protein quality scoring comes in.
The older metric, PDCAAS (Protein Digestibility-Corrected Amino Acid Score), measures the limiting amino acid in a protein and adjusts for how well the whole protein is digested. The newer DIAAS (Digestible Indispensable Amino Acid Score) goes further by measuring the digestibility of each essential amino acid individually at the end of the small intestine, which is where absorption happens. When both scores are calculated for the same foods, PDCAAS tends to give higher values for plant proteins and processed proteins like soy protein isolate, potentially overestimating their quality.17PubMed. Values for digestible indispensable amino acid scores (DIAAS) for some dairy and plant proteins may better describe protein quality than values calculated using the concept for protein digestibility-corrected amino acid scores (PDCAAS)
For everyday purposes, this means that 30 grams of protein from eggs, dairy, or meat delivers more usable amino acids than 30 grams from beans, lentils, or wheat. The gap narrows when plant proteins are combined to complement each other’s limiting amino acids, which is the traditional logic behind pairing rice with beans. But the label does not capture any of this. It just says “protein.”
Antinutrients and the Bioavailability Gap
Plant foods contain naturally occurring compounds called antinutritional factors that can reduce how much protein your body absorbs. Trypsin inhibitors in soybeans and kidney beans can cut protein digestibility by up to half in some experimental settings, and phytates in cereals and legumes can reduce amino acid digestibility by up to 10 percent.18PubMed. Impact of antinutritional factors in food proteins on the digestibility of protein and the bioavailability of amino acids and on protein quality Cooking, soaking, fermenting, and sprouting reduce these compounds substantially, which is one reason traditional food preparation methods evolved the way they did.
These same antinutritional factors also have potential health benefits at lower doses, including antioxidant and anti-inflammatory effects.19PubMed Central. Decoding the Duality of Antinutrients: Assessing the Impact of Protein Extraction Methods on Plant-Based Protein Sources So the goal is not to eliminate them entirely but to be aware that raw or lightly processed plant foods may deliver less usable protein than their labels suggest. If you rely heavily on plant protein and want to maximize absorption, proper cooking and preparation matter more than obsessing over label grams.
Edible Insects and Other Novel Proteins
As alternative protein sources gain attention, the same measurement problems that affect traditional foods show up in new forms. Edible insects, for example, contain chitin, a nitrogen-rich fiber that inflates crude protein values when measured by standard nitrogen methods. A study of five edible insect species found that their actual nitrogen-to-protein conversion factors ranged from 4.93 to 5.76, all lower than the standard 6.25.20PubMed. Low protein digestibility-corrected amino acid score and net nitrogen-to-protein conversion factor value of edible insects Using 6.25 on cricket flour, for instance, overstates its protein content meaningfully. The same study also found low PDCAAS values for several insect species, suggesting that the quality of insect protein, not just the quantity, may be lower than marketing materials imply.
Cultivated meat, algal proteins, and fungal mycoprotein all face similar measurement challenges. Each has a unique amino acid profile and varying amounts of non-protein nitrogen. As these products enter the mainstream food supply, existing conversion factors may need to be recalibrated. For now, if you see protein claims on novel food products, it is reasonable to assume they use the same crude-nitrogen methodology as conventional foods and carry the same inherent overestimation.
Practical Steps for Better Protein Estimates
Given all the uncertainty built into protein measurement, here is what actually helps when you are trying to track your intake:
- Use a food scale: Weighing ingredients before cooking eliminates the biggest source of error for most people, which is guessing portion sizes.
- Match your database entry to your prep method: If you weighed chicken raw, search for raw chicken. If you weighed it cooked, search for the cooked entry. Mixing these up introduces a 15 to 25 percent error.
- Stick to one tracking app: Switching between apps introduces inconsistency. Research suggests most popular apps underestimate protein by roughly 10 to 12 grams per day relative to research-grade software, but the bias is fairly consistent within a single app, so trends over time remain meaningful.
- Be skeptical of novel-food protein claims: Products made from insects, algae, or highly processed plant blends often use the 6.25 conversion factor, which likely overstates their protein content more than it does for meat or dairy.
- Do not ignore preparation: Soaking, cooking, and fermenting legumes and grains reduces antinutritional factors and improves actual protein absorption, even though the label number stays the same.
When the Label Gets It Wrong on Purpose
The gap between crude protein and true protein is not just an academic issue. It has been exploited deliberately. In 2008, melamine was found added to infant formula and pet food in China because melamine is 67 percent nitrogen by weight, and even tiny amounts dramatically inflate crude protein readings. The scandal killed six infants and sickened hundreds of thousands of children. The Kjeldahl method could not distinguish melamine nitrogen from protein nitrogen. Laboratories now screen for melamine specifically, but the episode exposed how easy it is to game a measurement system that equates nitrogen with protein.
More mundane forms of inflation happen legally. Sports nutrition products are a frequent offender. A study that tested high-protein sports foods against their label claims by performing Kjeldahl analysis found that while most products’ crude protein matched the label reasonably well, the underlying method still counted non-protein nitrogen as protein.21PubMed Central. Authentication of protein, fat, carbohydrates, and total energy in commercialized high protein sports foods with their labeling data Some manufacturers add free amino acids like glycine or taurine, which boost nitrogen readings without contributing the full spectrum of essential amino acids you need. This practice, sometimes called “amino spiking,” is not illegal in most jurisdictions but means the functional protein content is lower than what the label promises.