Horse feeding plan (US): what the law asks, the numbers in pounds, and the hazards in the hay

By Sebastien Kister

Last reviewed 12 October 2026

No law in the United States says how much a horse should eat. State cruelty statutes ask for food that is "necessary" or "proper", and leave the amount to whoever is judging the horse — which in practice means its body condition. This page gives the numbers US extension services and the AAEP publish, in pounds, the limits the veterinary consensus statements set, and the hazards in American hay that a European guide never has to mention.

What US law asks: necessary, proper, sufficient — and never a number

There is no federal feeding standard for a privately kept horse. The federal Animal Welfare Act defines "animal" so that it "excludes … horses not used for research purposes". What applies is the animal-cruelty statute of the state the horse is in, and they read alike:

None of them puts a number on it: no pounds of hay, no hours between feeds, no percentage of bodyweight. The closest the US comes is in two documents that score the horse rather than the feed. New Jersey's equine care regulation, which binds, requires every horse to have "a BCS of at least a level 2.0" — on the 0 to 5 Carroll and Huntington scale, not Henneke's. And the UC Davis minimum standards written for California animal control officers, which are guidance rather than regulation, call a Henneke score of 3 out of 9 "the minimum standard", and ask for confined horses to be fed "at least once, but preferably twice, daily", at "one and a half to three percent of their body weight per day".

The AAEP's care guidelines for rescue and retirement facilities set the target higher: a body condition score "between 4 and 6 points on the Henneke condition score chart", about 1.5–2% of bodyweight a day as dry matter, and feeding "at least twice daily" for a confined horse. Those are the numbers worth planning to; the statutes are what is left when nobody did.

How much: the numbers in pounds

• Forage. Extension services put it at about 1.5–2.5% of bodyweight a day, and the University of Minnesota says most horses should eat 2% of their bodyweight in hay — 20 lb for a 1,000 lb horse, with an absolute floor around 1–1.25% for a horse on a weight-loss plan under a vet.
• Water. Most 1,000 lb horses drink 10–12 gallons a day, 12–16 when working in summer heat (University of Minnesota). The AAEP gives a range of 5–20 gallons and says a horse "should always have unlimited access to fresh water".
• Grain. No more than about 0.3–0.4% of bodyweight in one feeding, which is 3–4 lb for a 1,000 lb horse — and the AAEP's wording is the one to keep: grain "should always be measured by weight rather than volume (e.g., 1 pound, rather than 1 'coffee can')".

The same goes for hay. A flake is not a unit: Penn State weighed flakes at around 5 lb in one cutting and 7 lb in another, with weights varying inside one bale, and Mississippi State puts grass hay flakes at 2–5 lb and alfalfa at 3–8 lb. "Two flakes" can be 8 lb or 20. Weigh a few once, with a hanging scale, and the plan can say pounds.

The research is in kilograms, the barn is in pounds

The veterinary consensus statements below give their limits in grams per kilogram of the horse's bodyweight, so none of them can be applied to "half a scoop" without knowing what the horse weighs and what the scoop holds. The conversion is 2.2 lb to the kilogram: a 1,000 lb horse is about 450 kg, and 1 g per kilogram of bodyweight works out at almost exactly 1 lb per 1,000 lb of horse — which makes the starch limits below easy to carry in your head.

A scoop is a volume, and feeds differ in density — the same scoop can hold noticeably different weights of two different products, and a heaped one differs from a level one. Weigh the scoop of each feed once, on a kitchen scale, and write the number on the bin. That single measurement turns every recommendation on this page from a slogan into something you can check.

Starch: the two numbers with a figure attached

The ECEIM consensus statement on equine gastric ulcer syndrome gives an unusually concrete dietary recommendation:

The diet should not exceed 2 g/kg bodyweight of starch intake per day or more than 1 g/kg bodyweight of starch per meal.

For a 1,000 lb horse that is about 2 lb of starch a day, and 1 lb in any one meal. The same statement reports that exceeding 2 g/kg bodyweight per day is associated with roughly a two-fold increase in the likelihood of squamous gastric disease of grade 2/5 or worse.

Note what is being limited: starch, not feed. Four pounds of a 25% starch feed is 1 lb of starch — the whole per-meal limit for a 1,000 lb horse — and the same four pounds of a 12% feed is half of it. The starch content is on the bag or available from the manufacturer, and without it the limit cannot be applied. This is also why splitting the same daily quantity across more meals is not a cosmetic change — the per-meal figure is a separate limit from the daily one.

Where the evidence is weaker than the advice

"Feed more forage and the ulcers will sort themselves out" is the most repeated claim in this subject. The consensus statement does not support it as stated. It records that free access to fibrous feed is widely considered to reduce the risk of gastric ulceration "although strong evidence supporting this belief is also lacking", and that in a study comparing a high-fiber diet against an iso-energetic low-fiber one, both the number and the severity of squamous lesions were greater in the high-fiber group. On turnout, it calls the evidence conflicting.

None of that makes forage unimportant — it remains the basis of the ration, and straw fed as the only forage is associated with an increased likelihood of squamous disease. What it means is narrower and more useful: forage alone is not a treatment, and a horse with suspected ulcers needs a vet and a scope, not a bigger hay net and another six weeks.

A page that tells you only the comfortable half of the literature is not worth trusting on the rest of it.

The easy keeper: the numbers for weight loss

For obesity and insulin dysregulation, the ECEIM consensus statement on equine metabolic syndrome recommends a mixed-species grass hay-based diet at 1.25–1.5% of actual bodyweight as dry matter intake, equivalent to 1.4–1.7% as fed. It is candid that there is debate about the optimal degree of restriction, "owing to concerns about not meeting minimum roughage requirements" — which is precisely why this is a number to apply with a vet rather than alone.

Two findings from the same statement are worth knowing before you plan the routine:

• Soaking works, but it is not precise. Soaking hay for 7–16 hours at ambient temperature reduces water-soluble carbohydrates by 24–43% and is associated with more than double the loss of body mass compared with the equivalent dry hay. The spread is wide because hay varies — soaking lowers the sugar, it does not standardize it.
• Restricting turnout hours often does not restrict intake. Horses compensate by eating faster: over 0.9% of bodyweight as dry matter has been recorded in just three hours at pasture. Grazing muzzles cut intake to 77–83% of unmuzzled grazing, while keeping the horse foraging and moving.

Cereal-based feeds, fruit and vegetable treats are excluded from these diets for the same reason: their non-structural carbohydrate content.

Body condition score: the number that says whether it worked

Everything above is an input — what goes in the bucket, how much of it, how often. Condition score is the output: a hands-on estimate of fat cover at set points on the horse, and the only thing that answers the question the ration was chosen to answer. Is this working?

It exists because the eye is unreliable, and not vaguely so. The ECEIM metabolic syndrome statement records that a tendency for owners of obese horses to underestimate their horse's condition has been identified by several studies, and sets an uncomfortable explanation beside the usual ones: equine obesity is now prevalent enough that it "can lead to habituation and acceptance of the status as apparently normal". The yard you keep your horse on has calibrated your eye, and not necessarily well.

Two scales, and the number means nothing without knowing which

• 1 to 9 — Henneke et al., 1983. 1 is emaciated, 9 is extremely fat. This is the scale most published research works in, the ECEIM metabolic syndrome statement included. It was developed at Texas A&M on Quarter Horse broodmares, and it is the one US law enforcement uses: Tarleton State notes that it is "used by law enforcement agencies as an objective method" in cruelty cases.
• 0 to 5 — Carroll and Huntington, 1988. 0 is emaciated, 5 is extremely fat, and scores of 2 and 3 describe moderate and good condition. This is the scale in general use in the UK and much of Europe — and, unusually, the one New Jersey's regulation adopts.

So a condition score of 3 is a thin horse on one scale and a good one on the other. A score written down without its scale is not information: record it as 3/5 or 3/9, never as 3. It is exactly the sort of detail that survives one conversation and is gone by the next.

The neck is scored separately

Fat over the crest of the neck gets a number of its own, because it does not track the rest of the horse and it does not mean the same thing. The cresty neck score runs from 0 (no crest) to 5 (a large crest that drops to one side), with 3 or more taken to indicate one. Carter et al., who described it, reported it "useful in the assessment of neck crest adiposity" and possessing "physiological relevance, as demonstrated by associations with blood variables" — the neck measurements it tracks were the ones that correlated with insulin. This is why it is scored apart: the ECEIM statement defines obesity to include regionally excessive fat, and reports regional adiposity on that cresty-neck definition in 33% of horses surveyed in southwest England. A horse can carry an unremarkable overall score and still have a crest worth mentioning to a vet.

Where scoring is weaker than it sounds

The same statement is candid about the limits of the tool it recommends, and each of these changes what you may conclude from a number:

• It flattens out at the top. The relationship between total body fat and condition score is nonlinear "when BCS is greater than 7/9" — past that, the score stops tracking the fat, and body mass is the better measure.
• It can fail to move while the diet is working. The statement reports "variable reported changes in BCS after 6-12 weeks of energy restriction". A score that has not shifted in two months is not by itself evidence of failure.
• The first week is largely not fat. Larger losses are expected in the first week of restriction "because of reduced gut fill" once intake drops.
• It can stall after twelve weeks. Beyond that the same restriction may stop having the same effect, because resting metabolic rate adapts downward. A plateau is a reason to go back to the vet, not to quietly cut further.

Which is why the answer is both numbers, and neither by eye. For body mass the statement calls a weighbridge ideal, naming weight tapes and the apps that combine belly, girth and rump measurements as useful alternatives; the score is what tells you where that mass sits. Estimating the weight by looking is the one method the literature is unambiguous about, and it is not a good one.

Three hazards in American feed

• Blister beetles in alfalfa. Common in the southern and eastern US, they carry cantharidin, and a few dozen to a few hundred beetles — depending on the species — can kill a 1,000 lb horse. They are trapped in the hay when it is cut and crimped in one pass. Oklahoma State's advice to buyers: hay "harvested before May or after September", since adults typically emerge after the first cutting, and a look inside each flake. Infested hay stays toxic however long it is stored.
• Fescue and the pregnant mare. Most of the roughly 35 million acres of tall fescue in the US carries an endophyte that causes prolonged gestation, a thickened placenta ("red bag") and no milk. The University of Kentucky's advice is to "remove pregnant mares from infected forage 60-90 days prior to foaling" — diluting it with other feed does not work.
• Ionophores in cattle feed. Monensin and lasalocid are routine in cattle and poultry feed and lethal to horses; federal labeling says not to let horses near feed containing them. Contamination at mills that also mix horse feed is the real risk — in August 2024 more than 70 horses died in Elk City, Oklahoma, after monensin was mixed into horse feed at the mill. Buy horse feed from a mill that does not run medicated cattle feed on the same line, or ask.

And one American habit worth borrowing: testing the hay. Extension services are blunt that "laboratory testing is the only way to know if hay is low in sugar"; for a metabolic horse, the target Oregon State cites is under 10% non-structural carbohydrate.

Write it down, because you are not the one feeding

Every number above is defeated by the same thing: the ration in someone's head. The person who mixes the feed on a Saturday is often not the person who decided it, the "one scoop" that was agreed in March drifts by August, and a change made for a good reason on Tuesday is invisible to whoever is there on Wednesday.

Write the ration down where everyone who feeds can read it — feed, amount, when, and which days, because the grain that stops on the weekend is exactly the detail that gets lost. Keep the starch figure for each product beside it, so the per-meal limit can actually be checked rather than estimated.

Sources

• Sykes et al. — European College of Equine Internal Medicine Consensus Statement: Equine Gastric Ulcer Syndrome in Adult Horses, Journal of Veterinary Internal Medicine 29(5):1288 (starch per meal and per day, forage and turnout evidence, straw)
• Durham et al. — ECEIM consensus statement on equine metabolic syndrome, Journal of Veterinary Internal Medicine 33(2):335 (restricted intake rates, hay soaking, grazing muzzles, NSC, owner underestimation, the limits of condition scoring)
• Henneke et al. — Relationship between condition score, physical measurements and body fat percentage in mares, Equine Veterinary Journal 15(4):371 (the 1 to 9 condition score)
• Carroll and Huntington — Body condition scoring and weight estimation of horses, Equine Veterinary Journal 20(1):41 (the 0 to 5 condition score)
• Carter et al. — Apparent adiposity assessed by standardised scoring systems and morphometric measurements in horses and ponies, The Veterinary Journal 179(2):204 (the cresty neck score)
• 7 U.S.C. §2132(g) (the Animal Welfare Act's definition of animal, which excludes horses not used for research)
• State statutes: Texas Penal Code ch. 42, California Penal Code §597, New York AML §353 and §356, Florida §828.12 and §828.13, Kentucky KRS 525.130 (2017 Act), Oklahoma 21 O.S. §1685 (as reproduced in SB 945)
• N.J.A.C. 2:8-3.2 (New Jersey: a body condition score of at least 2.0 on the 0 to 5 scale)
• UC Davis Center for Equine Health — Minimum Standards of Horse Care in the State of California (2023; guidance for animal control officers)
• AAEP — Care Guidelines for Equine Rescue and Retirement Facilities (condition score 4 to 6, intake, feeding frequency, weighing grain, water)
• Tarleton State University — Henneke body condition scoring (its Texas A&M origin and use in cruelty cases)
• University of Minnesota Extension — Ten things you should know about feeding the mature horse and Using hay in horse diets (forage, grain per feeding, water)
• Penn State Extension — The importance of weighing horse feed and Mississippi State Extension P3490 (flake weights)
• Oklahoma State EPP-2072 — Blister beetles and alfalfa and New Mexico State CR-536 (cantharidin, cutting dates, lethal dose)
• University of Kentucky ID-144 — fescue toxicosis in horses and UK Ag Equine Programs (effects on mares, 60–90 days, acreage)
• 21 CFR 558.355 and 558.311 (monensin and lasalocid labeling), University of Arizona az2090 — ionophore poisoning (the 2024 Elk City deaths)
• Oregon State EM 9415 — sugar and NSC in equine pasture and hay (hay testing, under 10% NSC)
• National Research Council — Nutrient Requirements of Horses, 6th revised edition (the reference work underlying most published rations)

Cited by DOI rather than by publisher link on purpose: both consensus statements have already moved publisher since publication, and the DOIs survived the move. Consensus statements are revised; check for a newer edition before relying on one.

Also available in English (United Kingdom, Australia) · Español (España, México) · Português (Portugal, Brasil) · Deutsch (Deutschland, Österreich, Schweiz) · Français (France, Belgique, Suisse) · Italiano (Svizzera, Italia) · Nederlands (Vlaanderen, Netherlands) — one question answered fifteen ways, two of them Belgian and not agreeing with each other. Germany counts in hours and Austria in feeds a day; the US states, England, Switzerland, Spain, Portugal, Mexico and France all stop at sufficiency without a figure, and Brazil's criminal law does not mention food at all — its one national figure, four hours of work without water or food, sits in a veterinary council's resolution. Two documents pick a condition scale: Italy's code names a target score, and the Australian Capital Territory's mandatory code a floor and a ceiling. Wallonia is the outlier in both directions: the most precise numbers of the set — 16 hours of access, no more than 4 without food — sitting in an advisory opinion that binds nobody, while the code that does bind says only that the feed must suit the animal.

Next: what belongs in a horse's health record — and who should be able to see it.