Working Dog Nutritional Requirements: What the Standards Actually Guarantee

There is a sentence on the Association of American Feed Control Officials’ own consumer page that almost nobody quotes. It says that “a service, hunting or working dog or a pregnant or nursing pet may not receive enough nutrition from a food designed for a sedentary adult house pet.”

That is the organization whose model language governs pet food labeling in the United States, telling you in plain English that a service dog may need more than an ordinary bag provides.

Now here is the part that matters. AAFCO has exactly two nutrient profiles for dogs: Growth and Reproduction, and Adult Maintenance. Neither of them tells you how much of it to put in the bowl, which is a body condition question rather than a label question. There is no working dog profile. No performance profile. No athletic profile. The regulator’s own model language names your dog’s job, admits the need, and then offers no standard to meet it.

That gap is where the marketing lives. “Performance,” “working formula,” “sport,” “high energy” — on a bag, legally, those are adult maintenance foods with an adjective on the front.

This article is about what a sound dog food would actually contain, what the rules genuinely require, and the distance between those two things. It names no product and recommends no brand. That is deliberate: the standard has to be written before and independently of any product, or it is not a standard at all.

The short version

Nobody approves a dog food formula before it is sold. “Complete and balanced” means one of two things: a spreadsheet calculation that no animal ever ate, or a twenty-six week trial that can finish with six dogs and four blood tests, and passes so long as no dog loses more than fifteen percent of its body weight.

The vitamins that make a food “complete” are not in the meat and rice on the label. They are synthetic compounds bought in from a supply chain concentrated almost entirely in one country, and the federal rule permits a manufacturer to verify them by reading a certificate rather than testing the drum. When that has failed, it has failed by thirty-fold and sixty-fold, not by percentages.

And when researchers actually put sixty-one foods through a laboratory in 2025, every one of the twenty-three mass-market bags passed, and forty percent of the boutique ones failed.

The one dietary intervention ever proven to extend a dog’s life is not in any bag, costs nothing, and is therefore advertised by no one. It is at the end of this article.

Who actually sets the rules, and what they do not promise

AAFCO is not a regulator

This is the single most misunderstood fact in the subject, and AAFCO says it in its own words: “AAFCO itself has no regulatory authority whatsoever.” The FDA says the same thing independently, describing AAFCO as a voluntary association of local, state and federal feed control officials.

AAFCO writes model language. States choose whether to adopt it into their own law. Membership is limited to government feed control employees, and industry participates as non-voting advisors.

So when a bag, a blog or a salesperson says a food is “AAFCO approved,” they are describing something that does not exist. AAFCO is blunt about it: “AAFCO does not approve, certify or otherwise endorse pet foods.” The same goes for “WSAVA approved” — the World Small Animal Veterinary Association states plainly that its Global Nutrition Committee “does not endorse, approve, recommend or support specific products or companies.” A bag or a blog claiming either is describing a self-assessment against a questionnaire, not a certification.

What the FDA does and does not do

The FDA requires that animal food be safe to eat, produced under sanitary conditions, free of harmful substances and truthfully labeled. But the agency states directly that food for animals “doesn’t need FDA’s pre-market review to be legally marketed” unless a specific substance is a food additive.

The Food Safety Modernization Act added 21 CFR Part 507, preventive controls for animal food. Read that rule precisely, because a great deal of marketing depends on people not doing so: Part 507 governs safety and sanitation processes. It does not govern nutritional adequacy. Nothing in it requires a manufacturer to demonstrate that a food sold as complete and balanced actually meets a nutrient profile.

There is one more piece of housekeeping worth knowing. The Memorandum of Understanding between the FDA and AAFCO expired on 1 October 2024. It was replaced not with a rule but with two guidance documents, one of which describes itself as “an interim process providing an additional way for firms to consult with the FDA.” A completed consultation is not an approval.

What does “complete and balanced” actually mean on a dog food bag?

It means the food met one of two AAFCO routes. Either it was formulated on paper to match a nutrient profile, in which case no animal ate it before the claim was made, or it passed a feeding trial. It is a floor, not a grade, and it says nothing about ingredient quality, digestibility or whether the finished bag was ever tested.

The strongest claim on a bag is six dogs and four blood tests

The feeding trial is the higher of the two routes. It is what a company means when it says its food was “tested in dogs,” and it is the claim people pay a premium for.

Here is the protocol, from AAFCO’s own published feeding protocols:

“A minimum of eight healthy adult dogs at least one year of age and of optimal body weight shall be required to start the test.”

“The test shall run for a minimum of 26 weeks.”

“A number of dogs, not to exceed 25% of those starting the test, may be removed for non-nutritional reasons or poor food intake.”

Twenty-five percent of eight is two. A trial can finish with six dogs.

“No individual dog shall lose more than 15% of its initial body weight.”

“Hemoglobin, packed cell volume, serum alkaline phosphatase and serum albumin shall be measured and recorded at the end of the test.”

Four blood values. That is the complete list. No taurine. No vitamin D status. No bone density. No cardiac imaging. No lifespan.

Six to eight adult dogs, six months, four blood tests, and a pass so long as no dog loses more than fifteen percent of its body weight. That is the most rigorous nutritional adequacy claim a bag of dog food can carry in the United States. What “complete and balanced” actually means goes through the full protocol, who checks it, and what happened when two teams measured whether foods meet the claim.

The FDA sets out both routes plainly. The other route — formulation — is a calculation run against a nutrient database. It is not fraudulent and it is not lazy; it is how most food is qualified, and a competent formulator with good data produces a good result. But it should be understood for what it is. AAFCO itself warns that “formulating a product according to the Profiles is only one part of a nutritionally sound, scientific development” and encourages manufacturers to run feeding trials anyway.

Is there an AAFCO standard for working dogs?

No. AAFCO publishes two dog profiles only: Growth and Reproduction, and Adult Maintenance. There is no working, performance, athletic or senior profile. A food labeled “performance” or “working formula” is legally an adult maintenance food. The words on the front describe the marketing, not a standard the food had to meet.

What happens when somebody actually measures the food

In December 2025, researchers published an analysis of sixty-one adult and senior dog diets from twenty-five brands in Frontiers in Veterinary Science. They ran the food through a laboratory and compared the results against the AAFCO adult maintenance requirements.

Eighteen percent — eleven of sixty-one — failed at least one requirement. Broken out by market tier, the result is the finding that should reorganize how you shop:

Seventeen individual failures across four minerals: potassium below the minimum, calcium above the maximum, phosphorus above the maximum, and calcium-to-phosphorus ratios above two to one.

Price did not predict compliance. It inverted it.

This is one exploratory study and it should not be treated as the last word. But it is directly measured, it is open access, and it points the opposite way from every assumption the premium tier is built on. A separate 2026 analysis of ninety-six products found that only 82.3 percent met their own label’s guaranteed analysis for the seven mandatory nutrients.

If you are disabled and on a fixed income, and you have been made to feel that you are letting your dog down by buying the cheaper bag, that is the number to hold onto.

Where the vitamins actually come from

This is the part of the subject that essentially nobody writes about, and it is the most important thing on this page.

The vitamins in your dog’s food are not extracted from the chicken and the sweet potato. They are synthetic or fermentation-derived compounds, manufactured industrially, blended into a premix by a specialist supplier, and dosed into the batch at fractions of a percent. Roughly eighty percent of the pet food industry buys premixes rather than making them.

Those compounds come from a very small number of places. A November 2025 supply chain assessment prepared for the Institute for Feed Education and Research, on the impact of vitamin and amino acid supply chain disruptions on US food security, put China’s share of global production capacity in 2023 as follows:

The same assessment reports that the United States relies on China for 78 percent of total vitamin imports. An earlier assessment put it more starkly: for biotin, folic acid and vitamin B12, there are no other sources.

This is not a claim about Chinese manufacturing quality. It is a claim about concentration. When one input has one source and every brand on the shelf buys from it, a single failure is not a brand problem. It is an industry problem.

The check is a piece of paper, and it has failed

The governing rule is 21 CFR 507.110. It permits a manufacturer to verify a supplier by onsite audit, or by sampling and testing, or by reviewing the supplier’s records, or by other risk-based activities. Testing the drum is one option among several. Reading the certificate that came with the drum satisfies the rule.

Here is what that has produced. All three of the following are quoted from the FDA’s own warning letters, which are public and searchable by the reference numbers given.

To Hill’s Pet Nutrition, November 2019 (letter 576564): “Your firm did not obtain Certificates of Analysis (COA) upon receipt of vitamin premix from your supplier,” and “As a result of your failure to follow your food safety plan, the hazard of vitamin D toxicity was not adequately managed at your receiving step.”

To Sunshine Mills, June 2019 (letter 576411): the firm used five shipments of an ingredient it was unaware contained a far higher concentration of vitamin D than the material it believed it was receiving — roughly sixty-seven times the intended potency, across five shipments, unnoticed.

To DSM Nutritional Products, October 2019 (letter 578300) — the premix supplier at the root of it. Two batches assayed at 2,903 percent of the amount claimed and 307 percent of claimed. The cause, in the FDA’s account: vitamin D and vitamin E drums of “similar appearance,” the wrong drum selected, no secondary review, and no clean-out between batches.

What reached dogs

The Hill’s vitamin D recall began on 31 January 2019 and expanded twice, to thirty-three varieties across eighty-six lots. Measured vitamin D levels ran from 100,170 to 107,282 IU per kilogram, against an AAFCO maximum of 3,000. That is roughly thirty-three to thirty-six times the legal maximum. The class action settled for $12.5 million with no admission of liability.

Be careful with what gets said next about that recall, because the internet is not careful. No FDA document publishes a confirmed count of dogs killed. Circulating figures trace to advocacy sites and litigation filings. Vitamin D toxicity is real and can cause kidney failure and death — the FDA says so plainly — but “reports to the FDA” and “confirmed deaths” are different things, and anyone who blurs them is selling you something.

It did not stop in 2019. Purina recalled a veterinary diet in early 2023 for “a supplier error resulting in potentially elevated levels of vitamin D.” Revival Animal Health recalled canine milk replacers in April and June 2026 for “variable levels of Vitamin D resulting in both low and elevated levels,” with two complaints of rickets — the same product line over-dosed and under-dosed.

Thiamine is the other recurring failure. A 2017 review counted six North American recalls since 2010, and noted that “These diets all had AAFCO nutritional adequacy claims.” In March 2026 the FDA advised that in a cat food it tested, “All eight lots tested by the FDA contained extremely low or no thiamine” — measured values including zero, against a minimum of 5.6 milligrams per kilogram.

Then there is aflatoxin, a mold toxin. The Sportmix recall by Midwestern Pet Foods began on 30 December 2020 and expanded to more than a thousand lot codes, exported to thirty-four countries. The FDA’s action level for aflatoxin in pet food is 20 parts per billion; the warning letter that followed found finished product as high as 558 parts per billion, about twenty-eight times the action level. The same letter found seven lots presumptive-positive for Salmonella that were distributed anyway.

The FDA reported more than 110 pets dead and more than 210 sick in connection with it, and stated explicitly that “Not all of these cases have been officially confirmed as aflatoxin poisoning.” Both halves of that sentence belong in any honest account. And for the record, no criminal prosecution followed; the consequences were a $6.375 million class settlement and a warning letter closed out in 2023.

And underneath all of it sits melamine. In 2007, melamine was substituted into wheat gluten and rice protein concentrate because it is full of nitrogen, and the standard protein test measures nitrogen. A veterinary laboratory survey documented 424 confirmed cases and 278 deaths. The lesson was not about one supplier. It was that the number on the guaranteed analysis can be faked, and was.

The pattern is the point

Look at those numbers together. Sixty-seven-fold. Twenty-nine-fold. Thirty-six times the maximum. Twenty-eight times the action level. Zero thiamine where 5.6 was required.

When this industry fails, it does not fail by a few percent. It fails by orders of magnitude — because the failure mode is not sloppy formulation, it is the wrong drum, the wrong certificate, the wrong ingredient entirely. And in almost every case above, the problem was found by pet owners and their veterinarians reporting sick animals, not by testing that caught it first. The long version is here: the dog food recalls, and what they have in common.

That is the honest answer to “is my dog’s food safe?” It is very probably fine. The system that makes it fine is not the one you think it is.

What gets lost between the mixer and the bowl

A food can be correctly formulated and still not deliver what the label says, because vitamins degrade. This is measured, and the numbers are larger than most people expect.

A Kansas State University thesis on nutrient stability through pet food manufacturing and storage found the following retention after processing and ambient storage:

Over twenty-four months, EPA fell from 125 to 78 milligrams per kilogram and DHA from 77 to 50. A peer-reviewed 2024 study corroborates the vitamin A picture, finding that most of the loss happens during pre-conditioning and the drying and extrusion steps.

This is why manufacturers add overages — more vitamin than the label claims — because a food formulated to hit its guarantee exactly would fail that guarantee well inside its shelf life. There is no published figure for a standard industry overage, and anyone quoting you one has invented it.

“Stressed storage” means a plastic bin in a hot garage. That is the practical part, and it is free. The FDA’s guidance is to store dry food below 80°F, in its original bag, with the top folded down. Tufts adds the reason the bag matters: manufacturers test shelf life in that packaging, exposure to air accelerates breakdown, and opened dry food should be used within two to three months. If you use a bin, put the whole bag inside the bin rather than pouring the food in loose.

Heat changes the protein too

Extrusion runs at 80 to 200°C. One consequence is the Maillard reaction — the same browning chemistry that makes toast — which binds lysine into a form the dog cannot absorb. A review in Nutrition Research Reviews found the gap between total and reactive lysine ran as high as 61.8 percent across products, averaging 15.4 percent in dog maintenance foods.

That matters because lysine is an essential amino acid and the guaranteed analysis cannot see the difference. The label counts the lysine. The dog cannot use all of it.

Rendering has a similar effect on protein quality. Measured against raw meals, rendered animal meals showed total amino acid digestibility of 79.2 to 84.8 percent versus 90.3 to 95.5 percent, and lower protein efficiency ratios. The conclusion was direct: “Rendered animal meals generally had lower protein quality than raw animal meals.”

None of this makes kibble dangerous. It makes the printed numbers softer than they look.

Minerals are not interchangeable either

The same named mineral can vary enormously by source. A 2022 review found zinc oxide sources ranging from 22 to 93 percent relative bioavailability — the identical ingredient on the label, differing four-fold in what the animal can actually absorb. Copper from copper oxide is described as unequivocally poorly available.

Organic and chelated forms are generally at least as available as inorganic ones, sometimes better — but the review is careful, and so should we be: most of that evidence comes from poultry, swine and cattle, not dogs. What is not in doubt is that “zinc” on an ingredient list tells you nothing about how much zinc the dog gets.

What the label can tell you, and what it cannot

Lisa Freeman, a board-certified veterinary nutritionist at Tufts, wrote the sentence that ought to be printed on the inside of every pet store: “The ingredient list is one of the most useless aspects of the pet food label.”

Here is why.

Ingredient order is pre-cooking weight, including water. Fresh chicken is about seventy percent water. A hundred grams of “deboned chicken” listed first becomes roughly thirty grams of dry matter in the finished kibble. A hundred grams of chicken meal listed third stays around ninety. As Tufts puts it, a diet with chicken first “may have less actual chicken than a diet that has chicken meal… as the second or third ingredient.”

Ingredient splitting is legal and undetectable. List peas, pea protein, pea starch and pea fiber separately and you get four small numbers where there was one large one — so each falls below the meat. The same trick works with corn: ground corn, corn gluten meal, corn bran. Nothing on the label lets you reassemble them.

Crude protein is a nitrogen assay. AAFCO says so directly: it “does not directly measure protein, but a component (nitrogen) of the protein instead.” That is precisely how melamine worked. A high protein guarantee tells you nothing about protein quality, amino acid balance, digestibility, or in the worst case whether the nitrogen came from protein at all.

Crude fiber understates true fiber by about half. Across fifty-one dog foods, total dietary fiber averaged 2.0 times the crude fiber figure. Using crude fiber to estimate calories inflated the result by around five percent — enough to drive roughly fifty-seven calories a day of overfeeding in a medium dog.

And mislabeling is not rare. A systematic review of eighteen studies concluded that “The mislabeling of pet foods appears rather common, even in those with ‘novel’ or ‘limited ingredients’ proposed for elimination diets,” with undeclared ingredients found in a median of forty-five percent of diets tested.

The one piece of label math worth learning

Guarantees are printed as-fed, so a ten-percent-moisture kibble and a seventy-eight-percent-moisture can cannot be compared directly. Convert both to dry matter:

Dry matter percentage = (as-fed percentage ÷ (100 − moisture percentage)) × 100

Worked through: a kibble at 26 percent protein and 10 percent moisture is 26 ÷ 90 × 100 = 28.9 percent on a dry matter basis. A canned food at 9 percent protein and 78 percent moisture is 9 ÷ 22 × 100 = 40.9 percent. The can wins comfortably, despite looking a third as protein-rich on the panel.

Better still, compare grams per 1,000 calories, which normalizes moisture and energy density at the same time. The calorie statement is mandatory on the label, and it must say whether the figure was calculated or measured.

“Fillers” — the word is pointed in the wrong direction

Everything above is the short version. The product name rules — why “beef,” “beef dinner,” “with beef” and “beef flavor” mean 95 percent, 25 percent, 3 percent and nothing at all — the direction each guarantee has to run in, and the two-list test for which ingredients genuinely contribute nothing, are all worked through properly in how to read a dog food label.

The standard line from the veterinary establishment is that fillers do not exist. WSAVA’s own FAQ says “Nothing in a pet food is truly ‘filler’ and every ingredient in a pet food must have a nutritional purpose.”

That overcorrects, and it is falsifiable. There is no legal definition of “filler,” true. But there are absolutely ingredients that add bulk and little else, and some of them have been measured doing exactly that.

The useful standard is not the word. It is the question: inexpensive ingredients that add no demonstrated nutritive value, used regularly to stretch the amount of food. That is testable. “Filler” is not.

The clean case: powdered cellulose

AAFCO defines powdered cellulose as “purified, mechanically disintegrated cellulose prepared by processing alpha cellulose obtained as a pulp from fibrous plant materials.” In practice the pulp is wood.

Measured in dogs against beet pulp, cellulose lowered whole-diet dry matter digestibility (77.2 versus 81.3 percent) and increased wet fecal output (254 versus 209 grams a day). In canine in-vitro fermentation, twelve-hour organic matter disappearance was 2.6 percent for cellulose against 38.6 percent for beet pulp, with short-chain fatty acid production of 0.04 versus 2.72 millimoles per gram — a sixty-eight-fold difference.

Then the finding that settles it. In a separate study, fecal short-chain fatty acids measured 251.1 micromoles per gram on cellulose against 267.0 on a no-fiber control. Cellulose produced less fermentation product than feeding no fiber at all, while adding roughly half again as much stool.

There is a legitimate narrow use — deliberate calorie dilution in weight-loss diets, where being calorie-free is the entire point. Beyond that, it is dilution by another name, and the claim that every ingredient has a nutritional purpose does not survive contact with it.

Pea fiber (9.6 percent fermentability), peanut hulls, oat hulls, rice hulls and potato product belong in the same conversation. Potato flour had the lowest ileal crude protein digestibility of six flours tested in dogs, at 64 percent. Soybean hulls do not belong in that list — at 479.7 micromoles per gram they behave like a real functional fiber, close to beet pulp.

The ingredients that were framed

Now the other half, because an article that only does the first half is just a different sales pitch.

Grain allergy, meanwhile, is rare. Of 297 dogs with confirmed food reactions, the culprits were beef 34 percent, dairy 17 percent, chicken 15 percent, wheat 13 percent, soy 6 percent and corn 4 percent — and food allergy affects only one to two percent of dogs overall. Corn therefore implicates roughly one dog in a thousand. The proteins that grain-free foods use more of are the top three on that list.

Here is the asymmetry, and it is the whole story. The word “filler” got attached to the ingredient with 93 percent digestibility, and never to the ingredient with 2.6 percent. A decade of marketing pointed the word at corn, and while everyone was busy avoiding corn, the ingredients that genuinely contribute nothing walked straight past — into, as it happens, the grain-free and boutique segment that sells hardest against corn. Whether grain-free is actually bad for dogs is its own question, and every study on it is graded here.

Are fillers in dog food actually bad?

The word is the problem, not the concept. There is no evidence that corn, wheat or soy in a properly formulated diet causes disease or shortens life. But powdered cellulose, pea fiber and hulls have been measured contributing almost nothing. Ask what an ingredient contributes and whether anyone can show you — not whether a marketer has labeled it a filler.

Why it happens: least-cost formulation

None of this is a conspiracy. It is arithmetic, and the industry describes it openly. Commercial feed is built by least-cost formulation — a linear programming problem in which software finds the cheapest combination of ingredients that clears every nutrient minimum. When an ingredient’s price rises, the optimizer substitutes a cheaper one that still clears the bar.

Read the objective function carefully: minimum cost, subject to nutrient minimums. Nothing in it rewards exceeding a minimum. A food built this way is, by construction, the cheapest possible thing that clears the floor — and the floor is a floor, not a target.

And the prices are public. From the USDA’s National Weekly Grain Co-Products Report of 17 August 2026: soybean hulls $105–130 a ton, corn around $140, meat and bone meal around $250, soybean meal $313.70–362.40. Structural fiber runs roughly a third to a half the price per ton of rendered protein meal. That is the incentive, in dollars, rather than asserted.

The words that cost you money and mean nothing

“Ancestral” and “biologically appropriate,” refuted by the genome

The claim is that dogs are wolves and should eat like wolves. In 2013 a team published in Nature the genomic signature of dog domestication, and found that ten genes with key roles in starch digestion and fat metabolism show signals of selection. The three headline loci — AMY2B for pancreatic amylase, MGAM for maltase-glucoamylase, and SGLT1 for intestinal glucose transport — describe an animal built to handle starch. The wolf premise also underpins raw feeding, which raises a separate question that has nothing to do with nutrition once the dog has public access rights.
Their conclusion: adaptations “allowing the early ancestors of modern dogs to thrive on a diet rich in starch, relative to the carnivorous diet of wolves, constituted a crucial step in the early domestication of dogs.”

Starch tolerance is a domestication trait, not a compromise. “Feed him like a wolf” is a claim about a species your dog is not.

To be fair in both directions: dogs have no dietary requirement for carbohydrate, and Merck says so. But no requirement is not the same as no use. They digest cooked starch at better than 95 percent and use it efficiently as an energy source that spares protein for other work.

Fresh subscription food

The gently-cooked subscription category is now around 17 to 21 percent of US dog food by value. The economics are worth seeing plainly: $6 to $15 a day, against roughly $2 a pound for kibble — three to five times per meal. Customer acquisition runs on a standardized fifty percent first-box discount backed by more than $100 million of advertising in a single year, against a site conversion rate of about 0.7 percent. Retention is structural, because dogs eat the same thing every day and the switching cost is high.

Higher digestibility and lower stool volume are real and reported. But no peer-reviewed randomized controlled trial shows superior health outcomes for commercial fresh or gently-cooked diets over complete-and-balanced extruded diets. That is not an attack on the category. It is the difference between a measured property and a health claim, and only one of those is being advertised.

Who owns the brand, and who is rating it

Two things get sold to you on the pet food aisle. The first is the idea that small, independent, premium brands are the honest alternative to the big companies. Here is who actually owns them.

Blue Buffalo built its entire identity on attacking the big pet food companies. It is General Mills. Orijen and Acana are Mars. Merrick is Purina. Two of the remaining “independent natural” names are owned not by a food company at all but by private equity firms, whose objective is a return inside a fund’s life.

You are frequently choosing between two products from the same parent, and paying a large premium for the difference in the bag design.

Two further facts sharpen it. Mars and Nestlé Purina together hold about 72 percent of economy dry dog food in the United States. And plant ownership varies enormously — the Pet Nutrition Alliance’s manufacturer survey found companies ranging from zero percent plant ownership, meaning everything is made under contract by someone else, up to 95 or 96 percent. The majority of manufacturers contacted did not respond at all, which is itself a finding. When many brands share one contract plant, a single failure spreads across brands that look unrelated on the shelf — and a brand owner that does not own the line may not control the fix.

There is a two-word test for this on most bags. “Manufactured by” means they own the plant. “Manufactured for” or “distributed by” means they do not.

And the site rating the food

The largest consumer pet food rating site in the English-speaking world is Dog Food Advisor. It was founded by Mike Sagman, a dentist. In January 2023 Wag! Group acquired it for $9 million, and reported $1.4 million and $1.6 million of revenue from it in the two following quarters, announcing “expanded partnerships with premium pet food brands.” That revenue is referral commission — a share of what you spend when you click through and buy.

Lisa Weeth, a board-certified veterinary nutritionist, put the qualification issue plainly: “I would not consider a human dentist who has read a lot of dog food packages qualified to assess” pet food quality.

But the credential is not really the point, and neither is the commission on its own. The structural problem is that the ratings are derived from what is printed on the label. Everything this article has covered — bioavailability, formulation competence, storage stability, quality control, whether the drum was ever tested — is invisible on a label. A star rating derived from reading a package is a rating of the packaging.

What that looks like in practice

Here is a concrete example, from a food we are currently running a long-term trial on ourselves.

The manufacturer’s own AAFCO nutritional adequacy statement on one formula in the line carries an explicit exception: it excludes the growth of large-size dogs. For anyone raising a large-breed service dog prospect, that is arguably the single most safety-relevant sentence printed on the bag — it is the manufacturer saying, in the regulated part of the label, that this food is not for a puppy who will mature over seventy pounds. That exception exists for a documented reason: the Merck Veterinary Manual states that excess dietary calcium above three percent on a dry matter basis “causes more severe clinical signs of osteochondrosis and decreased skeletal remodeling in young, rapidly growing large-breed dogs,” because large-breed puppies cannot down-regulate how much calcium they absorb. The damage is skeletal, and it is what ends working careers before they start. Feeding a large-breed puppy covers the ceiling, the growth rate, and the exact wording to look for on the bag.

The rating of that food did not reflect the exception.

Not because anyone was careless. Because of exactly the structural problem above: a rating built from the ingredient panel is not reading the adequacy statement, and the exception only exists in the adequacy statement. The ingredient list looked the same either way.

That is what a label-derived rating cannot do. It is not that it gets ingredients wrong. It is that the most consequential sentence on the bag is not an ingredient, so it is not in the model at all.

The reusable test: before you trust any rating of a dog food, ask what the rating was calculated from. If the answer is “the ingredient list,” you have been handed an assessment of the panel on the back, not of the food in the bag — and certainly not of whether it is safe for the dog you actually have.

And this is where we have to apply the same test to ourselves. Any site that reviews dog food, including this one, is subject to exactly the same question: who pays, and does the payment scale with the price of what you are persuaded to buy? That is why the standard on this page was written before any review, and why nothing on this page recommends a product.

Nobody is enforcing this

In November 2011 the FTC sent Blue Buffalo a closing letter after reviewing by-product, human-grade and antioxidant claims, closing without action once the claims were removed, and noting that “This action is not to be construed as a determination that a violation did not occur.”

No FTC enforcement action — as distinct from a closing letter — could be found against a major pet food manufacturer for unsupported health claims. The real consequences have come from competitors suing each other and from consumer class actions. This category is policed by its own rivals and by plaintiffs’ law firms, not by regulators.

That sentence is the whole article in miniature, and it is the reason the questions below matter more than anything printed on the front of a bag.

So does a service dog need a working dog food?

Almost certainly not, and the arithmetic is worth doing yourself.

The standard clinical calculation, from the Merck Veterinary Manual, starts with resting energy requirement: RER = 70 × (body weight in kilograms)^0.75. Adult maintenance is then roughly 1.8 × RER for an intact dog, 1.6 × RER for a neutered dog, and 1.4 × RER for a dog prone to obesity.

Then the caveat that belongs on the fridge: “Any given animal may require as much as 30% more or less of the calories calculated for its body weight than another animal with the same body weight.” The formula is a starting point, not an answer.

What does work actually cost? About 1.1 calories per kilogram of body weight per kilometer traveled. The published activity bands put agility, flyball and racing in the lowest tier, under a 25 percent increase over maintenance. Field trials and hunting sit in the middle. Sled racing and carting over ten miles are the top tier, above 100 percent — and sled dogs working in the cold can reach eight times maintenance.

A mobility or psychiatric service dog doing house, errands and appointments is nowhere near any of those bands. Energetically, she is an adult pet dog: Merck’s 1.6 to 1.8 × RER with a wide margin of error, adjusted by what you can feel over her ribs, not by her job title. A mobility service dog bracing and retrieving through a normal day is doing skilled work, not endurance work. Those are different things, and only one of them is a calorie problem.

Nobody has ever measured a service dog’s day

This is worth stating as the finding it is. Systematic searching turns up no published study measuring the activity level or energy expenditure of assistance dogs. No accelerometry, no doubly-labeled water, nothing — not for mobility dogs, psychiatric service dogs, guide dogs or medical alert dogs. Canine accelerometry studies exist for other populations. Not for this one.

The population being sold “working dog” food is the one population nobody has bothered to measure.

The closest available comparison is instructive in a different way. A study of sports dogs against pet dogs found median body condition scores of 5 versus 6, and that “78% of pet dogs were classified as ‘too heavy'” against 45 percent of sports dogs. Pet dog owners fed approximately sixty percent more per day.

Does my service dog need a performance or working dog formula?

For most service dogs, no. A mobility or psychiatric service dog’s daily workload is energetically similar to an active pet dog’s, and no AAFCO working dog profile exists for a “performance” food to meet. Feed a complete and balanced adult food to the dog’s actual body condition. Genuine endurance work — sledding, long-distance carting — is the exception.

The one real exception: scent and alert work

If your dog does medical alert or scent detection, there is measured evidence that diet affects performance. In a controlled crossover study of seventeen detection dogs, dogs were 1.42 times as likely to find a target on a high-polyunsaturated-fat diet than on a high-fat diet. The high-PUFA diet was the corn oil one, which is a pleasing detail given where this article started.

The same body of work found dogs were 1.49 times as likely to find a target before exercise as after it, and that dehydration of the nasal mucosa directly degrades odorant detection. For an alert dog handler the practical takeaway is measured and free: a hot, tired, dehydrated dog alerts worse.

On hydration in heat, there is one more useful finding. Detection dogs working in an average of 85°F offered a chicken-flavored oral electrolyte solution drank 15.61 milliliters per kilogram per hour, against 7.04 with plain water. Flavor drove the intake — not electrolyte content. That is the defensible reason to flavor the water, and it sits neatly beside the fact that electrolyte supplements themselves have never been associated with a proven performance benefit.

While we are there: supplements aimed at working dogs mostly do nothing, and two have been measured doing harm. Greyhounds ran slower on a gram a day of vitamin C, and slower on 1,000 IU a day of vitamin E.

Older and retiring dogs need more protein, not less

The old advice to restrict protein in senior dogs is outdated. The National Academies’ own consumer publication states that older dogs “appear to require somewhat more protein to maintain their protein reserves, perhaps as much as 50% more,” while needing about twenty percent fewer total calories.

Fewer calories and more protein means a higher protein density, not a “senior” bag chosen for the word on the front. This matters most for a dog coming to the end of her working life, where muscle loss and joint pain arrive together — which we covered in retiring a service dog.

One more thing worth knowing if you are managing an arthritic working dog. A 2022 systematic review and meta-analysis of enriched diets and nutraceuticals in canine osteoarthritis found omega-3 supplements had the highest weighted efficacy, while glucosamine and chondroitin showed a “very marked non-effect,” with 88.9 percent of trials showing no efficacy — the authors concluded they “should no longer be recommended for pain management.”

Glucosamine is the best-selling joint supplement in the pet aisle, sold to owners of exactly these dogs. The thing that does work is cheaper and barely marketed. Ask your own veterinarian for a dose; there is no verified canine milligram-per-kilogram figure worth printing here, and anyone who prints one confidently is guessing. The rest of the evidence, including the trial that measured dogs on a force plate instead of asking their owners, is worth reading before you spend anything.

The seven questions that replace reading the ingredient list

This is the reusable test, and it comes from the WSAVA Global Nutrition Committee’s guidelines, with three additions from Tufts. Email them to the manufacturer. The answers, and the willingness to answer at all, tell you more than the whole back panel.

1. Do you employ a full-time qualified nutritionist — a PhD in animal nutrition, or board certification by the ACVN or ECVCN?
2. Who formulates your diets, and what are their credentials?
3. Do you own your manufacturing plants? (“Made by” or “made for”?)
4. What quality control do you run on incoming ingredients and on finished product? Specifically: do you test the vitamin premix, or accept the certificate?
5. What research have you published in peer-reviewed journals?
6. What is the caloric content per cup and per kilogram?
7. Was adequacy established by feeding trial or by formulation?

Tufts adds a tell worth watching for: “Does the manufacturer bash other pet food companies in advertisements?” And a good practical one — will they give you the level of any nutrient on request, not just the seven on the guaranteed analysis? A company that can answer that has the data. A company that cannot, does not.

How can I tell if a dog food company knows what it is doing?

Ask whether they employ a full-time board-certified veterinary nutritionist, whether they own their plants, whether they test incoming vitamin premix or accept the supplier’s certificate, and whether adequacy came from a feeding trial or a formula. A company with good answers gives them readily. A company without them changes the subject to ingredients.

The one thing proven to add years, and nobody sells it

Everything above is about avoiding harm and seeing through claims. There is exactly one dietary intervention that has been proven to extend canine lifespan in a controlled trial, and it has nothing to do with what is in the bag.

Forty-eight Labrador Retrievers were followed from eight weeks old until death, in littermate pairs. One of each pair was fed twenty-five percent less than its sibling. Nothing else differed. The results, published across two decades:

“Major lifetime observations included 1·8 years longer median lifespan among diet-restricted dogs, with delayed onset of late life diseases, especially osteoarthritis.”

Median lifespan was 13.0 years against 11.2 — about fifteen percent more life.

And the joints:

“Median age at first identification of radiographic evidence of hip joint osteoarthritis was significantly lower in the control-fed group (6 years), compared with the restricted-fed group (12 years).”

For a working dog that is not a lifespan statistic. It is a career. Hip arthritis at six years versus twelve is the difference between a service dog retiring at seven and one still working at eleven — the difference between one dog and two, and between one training investment and two. It matters most in the large breeds that dominate service dog work, where hips are already the limiting factor.

Two honest caveats, because this site does not get to skip them for evidence it likes. The trial was forty-eight dogs of one breed at one kennel on one diet, and it was funded and conducted by Nestlé Purina. And “restricted” is a misleading word: the restricted dogs sat at a body condition score of 4.6 out of 9 and the control dogs at 6.7. The lean dogs were at a normal weight. The control dogs were the normal American dog. This is a study of leanness, not of deprivation.

The population data agrees. Across 50,787 neutered dogs in twelve breeds, overweight dogs had shorter median lifespans across the board — two and a half years in Yorkshire Terriers, around six months in Labradors.

How to actually do it

Use your hands, not the scale, and not the number on the bag. The WSAVA nine-point body condition scale defines the middle of the range like this:

• 4 of 9: ribs easily palpable with minimal fat covering, waist easily noted from above, abdominal tuck evident.
• 5 of 9: ribs palpable without excess fat covering, waist observed behind the ribs from above, abdomen tucked up from the side.
• 6 of 9: ribs palpable with slight excess fat covering, waist discernible from above but not prominent.
• 7 of 9: ribs palpable with difficulty, heavy fat cover.

Score muscle condition separately, by feeling over the spine, hips, shoulders and skull. This matters enormously for an older working dog, because you can have an overweight dog with moderate or severe muscle loss — and the fat hides it.

Two reasons to trust your hands over the label. First, owners get this wrong in a measurable direction: in a study of 174 owners, 51 percent scored body condition correctly and 38 percent underestimated it, and fewer than half interpreted a feeding label correctly. Second, the labels themselves run high. A Tufts analysis of eight dry adult foods found several recommended excessive calories for larger dogs, one by up to 61 percent. AAFCO’s own language calls feeding directions “guidelines that may need modification,” and notes that it does not standardize how they are calculated.

This is the whole argument in one line: the only intervention proven to add years to your dog’s life is free, and that is exactly why you have never seen it advertised. There is no margin in telling you to feed less. Every incentive in this industry runs the other way.

What this article is not

It is not a recommendation. No product is named as good or bad here, on purpose — the standard has to exist before the review does, or the review is just a sales page with footnotes. When we do review a food we bought, it gets measured against this page, and you can read exactly how we review things including what we spend and what we refuse.

It is also not veterinary advice. Your dog’s weight, joints, kidneys and gut are a clinical matter, and the single most useful thing in this article — feeding to a lean body condition — is worth doing with your veterinarian watching, especially in a growing large-breed puppy or a dog with an existing condition.

And if you are still deciding whether a service dog is the right path at all, that is a different question from what to feed one, and we answer it in do I qualify for a service dog.

What is the best dog food for a service dog?

There is no single answer, and any source that gives you one is selling something. Choose a complete and balanced food from a company that employs a qualified nutritionist, owns or closely controls its plants, and tests incoming ingredients — then feed it to a lean body condition. The amount matters more than the brand.

Does expensive dog food mean better dog food?

No. In the largest recent laboratory analysis, every one of the twenty-three mass-market foods tested met AAFCO adult maintenance requirements, while 40 percent of the boutique and premium foods failed at least one. Price did not predict quality — it inverted it. Larger manufacturers generally have more quality control staff and more published research, not less.