A close reading of the most popular statistic in the pro-meat internet, using the very study it cites.

Part of the same series as Two Acres, The Crop Deaths Argument, Counted Honestly, Rewilding the Plate, and The Pesticide Argument the Carnivore Movement Can’t Answer. Same rule throughout: steelman the other side, then do the arithmetic.

If you spend any time in online debates about meat, you have seen it: a yellow infographic with a donut chart declaring that 86% of global livestock feed is made of materials inedible by humans, with “only 14% edible by humans” stamped in a white circle. It circulates endlessly in regenerative-agriculture and carnivore communities as a one-image rebuttal to any criticism of animal agriculture’s efficiency.

Here is what makes this meme unusually effective: the number is real. It comes from a legitimate, peer-reviewed study led by researchers at the UN Food and Agriculture Organization, Mottet et al. (2017), published in Global Food Security [1]. The graphic even cites it correctly at the bottom.

But the same study that gave the meme its headline number also contains the numbers that quietly undo it: the net loss of human-edible food, the third of the world’s grain, the authors’ own call to reduce grain feeding. The debunk was, in effect, published alongside the claim in 2017. Nearly a decade later, the meme is still shared as though the case were settled in its favor.

A true statistic can still power a false argument, and this one does. Nearly every conclusion the meme invites you to draw is contradicted by the broader evidence, and, remarkably, several are contradicted by the cited study itself. What follows is a point-by-point breakdown, with sources throughout.

Where the 86% number comes from

What study is the 86% livestock feed statistic from?

The study is “Livestock: On our plates or eating at our table? A new analysis of the feed/food debate,” by Anne Mottet, Cees de Haan, Alessandra Falcucci, Giuseppe Tempio, Carolyn Opio, and Pierre Gerber, a team led from the FAO [1].

What is actually inside the 86%?

Livestock consume about 6 billion tonnes of feed (in dry matter) per year, including one third of global cereal production. Of that feed, 86% by dry weight consists of materials “that are currently not eaten by humans”: grass and leaves (46%), crop residues (19%), fodder crops (8%), oilseed cakes (5%), other by-products (5%), and other non-edible materials (3%). Human-edible grains make up 13% and other edibles 1% [1][2]. That is the meme’s donut chart, faithfully reproduced.

The paper was written partly to correct genuinely exaggerated claims, such as the old assertion that every kilogram of meat requires 10 or even 20 kilograms of grain. On that narrow point, it succeeded. The problem is everything the meme built on top of it. There are at least nine distinct issues.

Problem 1: Dry matter is the wrong yardstick

Is livestock feed measured by weight or by nutrition?

The 86% figure is a tally of feed by dry-matter weight. On a scale, a tonne of straw counts exactly the same as a tonne of grain, even though one is fibrous roughage and the other is calorie-dense and protein-dense food. Grass, crop residues, and silage are enormously bulky, so a weight-based accounting is structurally guaranteed to be dominated by them. The 86% is, in a real sense, an artifact of the unit chosen.

What happens when you count feed by calories and protein instead?

Change the denominator and the picture changes dramatically. Cassidy et al. (2013) found that feed crops make up 24% of global crop production by mass, but 36% of global crop calories and a remarkable 53% of global plant protein production, because the crops we feed to animals (maize, soy, oilseed meals) are precisely the densest ones [3]. Measured by the currencies that matter for food security, meaning calories, protein, or economic value, the human-relevant share of livestock feed is far higher than 14%.

Does “inedible to humans” mean it could never have been food?

Note the study’s own careful phrasing: materials “currently not eaten by humans” [1]. Edibility is partly a market classification, not a law of nature. Grain graded and handled as “feed-grade” could in many cases have been grown and handled as food; the category reflects economic choices made because livestock demand exists.

Problem 2: “Only 14%” is a mountain of food

How much human-edible food do livestock actually eat?

The word “only” is doing heroic work in that white circle. Livestock consume roughly 6 billion tonnes of feed dry matter per year [1], so the “mere” 14% that is human-edible works out to roughly 840 million tonnes annually, on the order of the entire global wheat harvest, which runs around 800 million tonnes per year. (The feed figure is dry matter and wheat is reported at harvest moisture, so treat this as an illustrative comparison rather than a precise like-for-like one. The point is the order of magnitude.)

Do livestock really eat a third of the world’s grain?

You do not have to take the arithmetic on faith, because the cited study states the same fact plainly: livestock consume one third of global cereal production [1]. A small percentage of a gigantic number is still a gigantic number.

How many people could be fed with the crops we feed to animals?

To put the opportunity cost in perspective: Cassidy et al. calculated that 36% of the world’s crop calories go to animal feed, and that shifting crops from feed (and biofuel) to direct human consumption could increase available food calories by as much as 70%, enough, in principle, to feed roughly 4 billion additional people [3]. That is a theoretical ceiling, not a policy proposal, but it shows the sheer scale of what “only 14%” represents.

Does this answer the “crops kill animals too” argument?

That acreage multiplier is also why the “crops kill field mice too” gotcha collapses on itself: routing crops through animals commissions more harvested acres per eater, not fewer. The full death-ledger accounting now runs to nine columns, and plant agriculture registers meaningfully in only one of them. The whole ledger is counted out here.

Problem 3: The cited study reports a net loss of human-edible food

How much human-edible feed does 1 kg of meat take?

This is the meme’s most damning problem: its own source contradicts its message. Mottet et al. calculate that producing 1 kg of boneless meat requires, on average, 2.8 kg of human-edible feed in ruminant systems and 3.2 kg in monogastric systems (pigs and poultry) [1]. Read that again: on average, the world’s meat production consumes roughly three times more human-edible food than it produces. Yes, these figures are far lower than the inflated “10 kg or 20 kg of grain” claims of the past, which was the paper’s corrective point, but a 2.8-to-1 or 3.2-to-1 losing ratio is still a losing ratio.

Are grazing cattle a net gain rather than a loss?

The exception the meme universalizes is grazing ruminants: cattle in grazing systems need only about 0.6 kg of human-edible protein to produce 1 kg of protein in milk and meat, making those specific systems net contributors of protein [2]. That is a genuinely interesting finding about one subset of production. It does not describe pigs, poultry, or grain-finished cattle, which is to say it does not describe most of the meat consumed in wealthy countries.

What did the study’s own authors actually conclude?

They call for reducing the use of human-edible grain in feed and improving feed-conversion ratios to prevent further expansion of arable land dedicated to feed production [1][2]. The paper is a case for reforming livestock production, not a vindication of the status quo. Lead author Anne Mottet has been explicit that the work argues for better livestock systems, with animals eating more residues and by-products and less grain, which is precisely the opposite of “nothing to see here.”

Problem 4: “Inedible by humans” is not the same as “costless to humans”

The meme’s implicit logic is that if humans cannot eat it, feeding it to animals is free. Land economics says otherwise.

How much cropland is used to grow animal feed?

Fodder crops, 8% of the donut, are not scavenged waste. They are crops deliberately cultivated on arable land that could grow human food. Overall, the study estimates livestock use about 40% of global arable land [2], and Poore and Nemecek’s global meta-analysis found that 38% of the world’s cropland is used for livestock feed, versus 50% for direct human food [4]. Roughly four of every ten cultivated acres on Earth are feeding animals, not people.

Can grazing land really not grow crops?

By the study’s own accounting, 57% of the land used for feed production is unsuitable for growing food, which means 43% is suitable. More specifically, the paper estimates that of the 2 billion hectares of grassland grazed by livestock, about 700 million hectares could be used as cropland [1][2]. That is an area roughly the size of Australia currently under pasture that could, physically, grow food.

What is the opportunity cost of land that cannot grow crops?

Land that cannot grow wheat can grow forests, store carbon, and host biodiversity. Poore and Nemecek estimate that a global shift away from animal products would reduce agricultural land use by about 75%, freeing an area the size of the US, China, the EU, and Australia combined, while still feeding the world [5][6]. And the carbon that freed land could pull back down is enormous: Hayek et al. (2021) estimate that letting vegetation regrow on land currently used for animal agriculture could sequester around 330 to 550 billion tonnes of CO2, on the order of the entire remaining carbon budget for a 1.5 °C target [18]. What that freed land becomes, and what actually happens to grazed ecosystems when the cattle come off, is the subject of Rewilding the Plate.

How much of the Amazon is cleared for cattle?

“Grass and leaves” in the donut includes pasture that was recently rainforest. The expansion of pasture for beef is the single largest driver of tropical deforestation, responsible for about 41% of it globally, around 2.1 million hectares per year (Pendrill et al. 2019) [7]. And cattle pasture occupies roughly 80% of the deforested land in the Brazilian Amazon [8]. When the meme’s green photograph of dewy grass includes ex-rainforest, “food we can’t eat” starts to look rather different.

Problem 5: The by-product sleight of hand

The donut classifies oilseed cakes (5%) and other by-products (5%) as leftovers that livestock heroically rescue. For some materials, such as brewers’ grains, beet pulp, and citrus pulp, that is fair. For the most important one, it is not.

Is soybean meal really just a by-product of oil?

Soybean meal (cake) is what remains after soybeans are crushed for oil. Calling it a “by-product” implies the oil is the point and the meal is a happy accident. In reality, the meal is most of the bean, about 44 of every 60 pounds crushed, roughly three-quarters of its mass [9]. For most of the crop’s history it was also most of the crush’s value; that value split has only recently shifted toward oil as the US renewable-diesel boom pushed soybean oil’s share past 50% [16]. Either way the meal is no afterthought. And here is the kicker: Mottet et al. themselves flag soybean cake production as a main driver of land use, even while counting it as only 4% of global feed intake [1]. The meme’s own source refuses the “innocent leftover” framing.

How much of the world’s soy goes to animal feed?

Zoom out to the whole soy economy and the picture sharpens: 77% of global soy is used as animal feed, while just 7% goes to direct human foods like tofu, tempeh, and soy milk. Of the total, over a third feeds poultry, about a fifth feeds pigs, and 6% goes to aquaculture [9]. Feed demand does not mop up the soy industry’s waste. Feed demand is the soy industry.

Are crop residues free to feed out?

Even genuine residues are not costless. Crop residues, 19% of the donut, have competing uses, chief among them being returned to the soil to maintain fertility and prevent erosion. “Inedible by humans” quietly conflates “no alternative use” with “no culinary use,” and those are very different claims.

One more cost hides in the feed complex: it is where the food system’s pesticide load concentrates, with federal residue tolerances on feed crops running hundreds to thousands of times those permitted on human produce. That regulatory paper trail is dissected here.

Problem 6: There is a third yardstick, and nobody counts it

Problem 1 showed that switching the denominator from dry weight to calories and protein transforms the picture. But calories and protein are not the only things a crop contains, and the third currency is the one this debate never counts at all.

Is animal feed just nutritionally empty bulk?

Feed crops are not inert filler. Maize bran carries one of the highest concentrations of ferulic acid reported in any plant tissue, and yellow maize is a documented source of lutein and zeaxanthin. Soybeans carry roughly 200 mg of isoflavones per 100 g of dry weight, plus saponins and a range of phenolic acids. Even the grass in the donut chart’s own photograph is real chemistry: fresh pasture runs on the order of 44 to 105 mg/kg of beta-carotene and 21 to 86 mg/kg of alpha-tocopherol on a dry-matter basis [19]. That is precisely the class of compounds, the carotenoids, tocopherols, and polyphenols, that no animal on Earth can synthesize and every animal must obtain by eating plants.

What happens to plant phytochemicals inside a cow?

Most of them do not survive the trip. Rumen microbes degrade a large share of the polyphenol load before it is ever absorbed; soy isoflavones, for instance, are broken down or converted to equol inside the rumen [20]. What does get absorbed is largely conjugated by the liver and excreted, with a substantial fraction of absorbed ferulic acid leaving in the urine [21], and reviews of the field conclude that tissue polyphenol content is not proportional to dietary intake [22].

The fat-soluble survivors are routed preferentially into liver, body fat, and milk fat rather than into skeletal muscle, which is the tissue people actually eat [23]. Cattle are in fact the best carotenoid depositors among common ruminants, which is why beef fat is yellower than the near-white fat of sheep and goats, and why grass-fed butter is yellower than grain-fed [24]. But muscle remains a poor depot. Where the transfer has been measured directly, the carry-over rates are small: studies tracking isoflavones from feed into milk report recoveries in the low fractions of a percent of intake [25].

How much antioxidant capacity actually survives into meat?

The largest survey of its kind assayed more than 3,100 foods. Animal-based foods came in at a median of 0.10 units of antioxidant capacity per 100 g against 0.88 for plant-based foods, a gap of roughly nine to one [26]. The means are far more lopsided still, 0.18 against 11.57, but the authors are candid that the plant mean is inflated by a minority of extreme products among the spices and herbal medicines, so the median is the fairer test. Their own like-for-like comparison is perhaps the cleanest statement of it: measured against the mean for meat and meat products, the fruit, nut, chocolate, and berry categories run from 5 to 33 times higher.

Two honest caveats belong here, because this argument should be held to the same standard as the rest of the post. First, that survey used an in-vitro chemical assay, and its authors say plainly that there is not necessarily a direct relationship between the antioxidant content of a food and the antioxidant activity that follows in a target cell. Antioxidant capacity scores are a measure of chemistry, not a measure of health outcomes. Second, nobody has ever directly assayed ferulic acid or isoflavone residues in beef muscle, so the honest claim is trace amounts by inference rather than a measured zero. The inference is well supported, resting on rumen degradation plus hepatic conjugation plus the measured preference for fat over muscle, but it is an inference, and worth labeling as one.

Does this mean animal foods have no nutritional advantages?

No, and the post should not pretend otherwise. Animal-source foods are genuinely good delivery vehicles for vitamin B12, heme iron, zinc, and preformed retinol, and in parts of the world where diets are narrow, that matters a great deal. Milk really does carry some equol through to the glass. The claim here is narrower and survives all of that: the phytochemical payload of the feed is not concentrated by the animal, it is largely dismantled by it.

Why this reframes the meme’s central move

“Inedible by humans” is offered as though the only question were whether the material could have been a meal. But the feed complex is full of chemistry that is valuable to a human body and is destroyed rather than delivered by routing it through a cow. The loss is not only caloric, and not only nutritional in the protein sense. It is a loss of the phytochemicals and the fiber alike, and almost none of it survives the trip. Whatever else the animal is doing, it is not concentrating this material for us.

Problem 7: A global average is being used to defend systems it does not describe

Do pigs and chickens eat grass?

The 46% “grass and leaves” figure is a global average, pulled upward by vast extensive and pastoralist systems, largely in lower-income countries, where cattle, goats, and sheep graze marginal land. Those systems are the ones the study’s favorable numbers (like the 0.6 protein ratio) describe.

They are not the systems that put meat on plates in the US or Europe. Industrial chicken, pork, and feedlot-finished beef, the overwhelming majority of rich-country meat, run on grain and soy, the exact human-edible and cropland-grown feeds the meme waves away. Recall that Mottet et al.’s conversion figure for monogastrics (pigs and poultry) is 3.2 kg of human-edible feed per kg of meat [1], and pigs and poultry are most of what is actually eaten.

Is grass-fed beef better for the climate?

Globally, grass-fed beef is a rounding error: the Oxford-led Grazed and Confused report found that fully grass-fed beef contributes roughly 1 gram of protein per person per day to the global supply, out of about 32 grams of animal protein consumed per person per day [10]. And a 2025 analysis in PNAS found that even under optimistic rangeland-sequestration assumptions, US grass-fed beef is no less carbon-intensive than industrial beef, and 3 to 40 times as carbon-intensive as a wide range of plant and animal alternatives, roughly 10-fold for common protein-dense foods, while cropland reallocated from beef feed to direct human food delivers 3 to 7 times more protein per unit of land, emissions, and nitrogen [11]. A pastoralist herd in the Sahel cannot vindicate a feedlot burger in Ohio, and it turns out a boutique grass-fed ranch cannot either. For the acre-by-acre arithmetic of what grazing land actually produces versus what cropland can, see Two Acres.

What about regenerative or adaptive multi-paddock grazing?

The strongest reply here is regenerative or adaptive multi-paddock (AMP) grazing, the claim that well-managed cattle build enough soil carbon to cancel their own emissions. It deserves a fair hearing, and the better field studies do find real gains: one southeastern-US comparison measured about 13% more soil carbon on AMP sites than on conventionally grazed ones (Mosier et al. 2021) [17]. But soil-carbon gains are finite, saturate within a few decades, reverse if the management lapses, and vary enormously by site. They are a one-time deposit, not a permanent subsidy. And the objection does not rescue the meme even on its best day, because the PNAS analysis above already builds in optimistic sequestration and grass-fed beef still finishes as the most carbon-intensive protein on the board. When regenerative-grazing advocates pressed exactly this point, the authors replied that such exceptions are real but far too rare to change the result [11].

Problem 8: Feed competition was never the whole efficiency argument

The meme frames the entire debate as a single question, “do livestock eat our food?”, and then declares victory on it. But the efficiency critique of animal agriculture rests on several independent pillars, and the feed/food question is only one.

Why are the conversion numbers this bad?

There is a reason these figures are as poor as they are, and it is not bad farming. The second law of thermodynamics guarantees that no energy transfer is free: some fraction always disperses as unusable heat. Ecologists reduced the consequence to a rule of thumb, that roughly a tenth of the energy at one level of a food chain survives into the next, and while that ten percent is an average rather than a constant (Lindeman’s original measurements ranged from a fraction of a percent up into the thirties), the direction never reverses [27]. A plant banks sunlight as calories. A cow spends most of those calories being a cow. What reaches a plate is the remainder.

This is why the efficiency critique cannot be managed away. Better genetics, adaptive multi-paddock rotation, holistic management: none of them repeal the law. They can argue about the size of the tax. They cannot argue about whether it is levied. Every figure in this section is that tax, scaled to a planet.

Two rebuttals are worth anticipating, because they sound like answers and are not. The first is that most of the loss is respiration, the animal breathing, moving, and staying warm. True, and irrelevant to the accounting: those calories still left the human food supply, and the fact that a cow burned them rather than a person does not put them back. The second is that the feed was inedible anyway, so the loss is free. Also false, as Problem 4 established: growing, harvesting, transporting, and feeding that material consumes land, water, fuel, and fertilizer that could have produced human food directly. Against that expenditure, the return includes not only the meat but the methane, the manure lagoons, and the nutrient runoff.

How much farmland does animal agriculture use?

Livestock use about 83% of the world’s farmland while supplying just 18% of calories and 37% of protein (Poore and Nemecek 2018, the largest meta-analysis of global food systems, covering about 38,700 farms in 119 countries) [4].

How efficient is beef at converting feed into food?

In the US, the average feed-to-food conversion efficiency across livestock categories is 7% to 8% for both calories and protein; for beef it is 3% on both metrics, meaning 97 of every 100 feed calories are lost. Merely reallocating US beef-feed cropland to poultry feed could meet the caloric needs of roughly 120 to 140 million additional people [12].

What share of greenhouse gas emissions comes from livestock?

Livestock account for roughly 12% of anthropogenic greenhouse-gas emissions, about 6.2 Gt CO2-equivalent per year, per FAO’s 2023 assessment, which revised down the older and still widely quoted 14.5% figure from FAO’s 2013 estimate [15]. And the Grazed and Confused team found that even under generous assumptions, soil carbon sequestration from grazing could offset only 20% to 60% of grazing cattle’s own emissions, leading them to conclude that expanding grass-fed ruminant production is “a self-defeating climate strategy” [10][13].

So even if the 86% statistic meant everything the meme wants it to mean, the land, climate, and conversion arguments would stand untouched. Rebutting one claim, and rebutting it badly, settles nothing.

Problem 9: Follow the meme’s logic and you arrive at much less meat

What if livestock only ate food humans cannot eat?

Here is the deepest irony. Suppose we take the meme’s principle completely seriously: livestock should exist to upcycle grass, residues, and true by-products, food we cannot eat, into food we can. Researchers have modeled exactly that world.

Van Zanten et al. (2018), reviewing the literature on livestock raised solely on “leftover streams” and grassland, found such systems could supply about 9 to 23 grams of protein per person per day, a real but modest fraction of the 50 to 60 grams humans need daily [14]. Ruminants fed solely on existing grasslands would supply around 7 grams per person per day [14]. Compare that with the roughly 32 grams of animal protein the average person on Earth already consumes daily [10], and with the far higher intakes typical in Western countries, and the conclusion is unavoidable: “livestock as upcyclers” is an argument for a livestock sector a fraction of the size of today’s, with rich-country meat consumption falling the furthest.

The meme borrows the moral glow of the upcycling model to defend a food system that does not remotely resemble it.

What the graphic gets right, and why it works

Intellectual honesty requires acknowledging what is true here, and it also explains the meme’s persuasive power.

The paper is real, correctly cited, and authored by credible FAO researchers. It genuinely corrected exaggerated feed-conversion claims that circulated for years. Ruminants really can convert cellulose humans cannot digest into nutrient-dense food, and specific grazing systems really are net protein contributors [1][2]. None of that is in dispute.

That is exactly what makes the meme effective: it is true but radically incomplete, the most durable species of misleading content. It answers a question almost nobody was asking (“is livestock feed mostly grain by dry weight?”) and presents the answer as though it resolved the questions people actually ask: whether animal agriculture competes with human food supply (it does, one third of the world’s grain), whether it uses land efficiently (it does not, 83% of farmland for 37% of protein), and whether current consumption levels are defensible on the meme’s own upcycling logic (they are not).

Consider the source

Who made the 86% livestock feed infographic?

The URL on the graphic, sacredcow.info, is not a research institution. It is the promotional website for Sacred Cow: The Case for (Better) Meat, a pro-meat advocacy book and companion film by Diana Rodgers and Robb Wolf. The graphic is, quite literally, marketing material.

Notice also the framing in the meme’s text: the efficiency critique is attributed to “alternative protein companies,” as though concern about feed and land use were a corporate talking point. In reality, the feed/food analysis comes from the FAO itself, the land-use figures from Oxford and Agroscope researchers publishing in Science, the conversion efficiencies from the Weizmann Institute publishing in Environmental Research Letters, and the climate analysis from an Oxford-led international team. The meme cites the academic establishment with one hand and recasts it as industry spin with the other.

The bottom line

The 86% statistic is real. The argument built on it fails on every front. By the very study the meme cites, livestock eat one third of the world’s grain and consume roughly three kilograms of human-edible food for every kilogram of meat they produce, and its authors call for grain feeding to be reduced. The “inedible” feed occupies land, including nearly 40% of the world’s cropland for feed, 700 million hectares of croppable pasture, and millions of hectares of former rainforest, that carries enormous opportunity costs. The “by-products” include a soy industry in which 77% of output feeds animals. And if you actually embrace the meme’s own principle, that livestock should eat only what we cannot, the science says that world contains a small fraction of the meat we currently produce.

Real number. Wrong conclusion. The strongest debunking of this meme was published in 2017, in the journal Global Food Security, by the authors the meme thanks in its own fine print.

Related reading in this series

• Two Acres: What Beef Produces vs. What Plants Can. The acre-by-acre math: roughly 500,000 calories of beef versus 20 to 32 million calories of plants from the same land, and why the corn and soy monoculture exists chiefly to feed animals.
• The Crop Deaths Argument, Counted Honestly. The field-mouse gotcha traced to its academic source, the same-year correction that flipped it five-to-one, and the nine columns of the death ledger the argument never counts: field deaths, slaughter, wild capture, predator control, disease culls, on-farm mortality, pesticide kills, dead zones, and habitat loss.
• Rewilding the Plate. What the freed land becomes: the regenerative-grazing story audited, and the peer-reviewed recovery that follows when the cattle come off.
• The Pesticide Argument the Carnivore Movement Can’t Answer. Feed crops carry the heaviest pesticide tolerances in federal law, and routing calories through animals multiplies the sprayed acreage your diet requires.

References

1. Mottet, A., de Haan, C., Falcucci, A., Tempio, G., Opio, C., and Gerber, P. (2017). Livestock: On our plates or eating at our table? A new analysis of the feed/food debate. Global Food Security, 14, 1–8. sciencedirect.com
2. Mottet, A., et al. (2017). More fuel for the food/feed debate. Feedipedia / FAO Broadening Horizons No. 47. The authors’ own open-access write-up, stating the 0.6 kg protein ratio for grazing cattle, about 40% of arable land used for feed, and about 700 million hectares of grazed grassland that could be cropland. feedipedia.org. Further summarized at whylivestockmatter.org
3. Cassidy, E. S., West, P. C., Gerber, J. S., and Foley, J. A. (2013). Redefining agricultural yields: from tonnes to people nourished per hectare. Environmental Research Letters, 8(3), 034015. iopscience.iop.org
4. Poore, J., and Nemecek, T. (2018). Reducing food’s environmental impacts through producers and consumers. Science, 360(6392), 987–992. (83% of farmland; 18% of calories; 37% of protein; cropland split of 50% food, 38% feed, 12% non-food per Supplementary Table S10.) science.org
5. Ritchie, H., and Roser, M. Half of the world’s habitable land is used for agriculture. Our World in Data. ourworldindata.org
6. Ritchie, H. If the world adopted a plant-based diet, we would reduce global agricultural land use from 4 to 1 billion hectares. Our World in Data. ourworldindata.org
7. Ritchie, H. Drivers of Deforestation. Our World in Data, summarizing Pendrill, F., et al. (2019), Global Environmental Change. (Beef equals 41% of tropical deforestation, about 2.1 million hectares per year.) ourworldindata.org
8. WWF. Unsustainable cattle ranching. (Cattle ranching accounts for about 80% of deforested land in the Brazilian Amazon; Nepstad et al. 2008.) wwf.panda.org
9. Ritchie, H. (2021). Soy. Our World in Data. (77% of global soy to animal feed; 7% to direct human food; poultry about 37%, pigs about 20%, aquaculture 6%. Meal is about 73% of the bean by mass, roughly 44 of every 60 lb crushed.) ourworldindata.org
10. Garnett, T., Godde, C., et al. (2017). Grazed and Confused? Food Climate Research Network, University of Oxford. (Grass-fed beef about 1 g protein per person per day globally; sequestration offsets at most 20% to 60% of grazing cattle emissions; “self-defeating climate strategy.”) oxfordmartin.ox.ac.uk. Coverage: carbonbrief.org
11. Eshel, G., Flamholz, A., Shepon, A., and Milo, R. (2025). US grass-fed beef is as carbon intensive as industrial beef and about 10-fold more intensive than common protein-dense alternatives. PNAS, 122, e2404329122. pnas.org. See also the authors’ reply to a regenerative-grazing critique, PNAS 122, e2506941122: pnas.org
12. Shepon, A., Eshel, G., Noor, E., and Milo, R. (2016). Energy and protein feed-to-food conversion efficiencies in the US and potential food security gains from dietary changes. Environmental Research Letters, 11(10), 105002. iopscience.iop.org
13. Inside Climate News (2017). Eating grass-fed beef isn’t as climate-friendly as you may think. insideclimatenews.org
14. Van Zanten, H. H. E., Herrero, M., Van Hal, O., Röös, E., Muller, A., Garnett, T., Gerber, P. J., Schader, C., and De Boer, I. J. M. (2018). Defining a land boundary for sustainable livestock consumption. Global Change Biology, 24(9), 4185–4194. onlinelibrary.wiley.com
15. FAO (2023). Pathways towards lower emissions: A global assessment of the greenhouse gas emissions and mitigation options from livestock agrifood systems. (Livestock agrifood systems about 6.2 Gt CO2-eq per year, about 12% of anthropogenic GHG emissions, revising down the older 14.5% figure from FAO 2013.) fao.org
16. CME Group (2025), “Soybean Oilshare,” and Janzen, J., and Wang, S. (2025), “The Soybean Industry Response to the Renewable Diesel Boom, Part 3,” farmdoc daily. (A 60-lb bushel yields about 44 lb meal and 11 lb oil; soybean oil’s share of crush value rose from about 25% to 35% up past 50% amid renewable-diesel demand.) cmegroup.com. farmdocdaily.illinois.edu
17. Mosier, S., Apfelbaum, S., Byck, P., Calderon, F., Teague, R., Thompson, R., and Cotrufo, M. F. (2021). Adaptive multi-paddock grazing enhances soil carbon and nitrogen stocks and stabilization through mineral association in southeastern US grazing lands. Journal of Environmental Management, 288, 112409. sciencedirect.com
18. Hayek, M. N., Harwatt, H., Ripple, W. J., and Mueller, N. D. (2021). The carbon opportunity cost of animal-sourced food production on land. Nature Sustainability, 4, 21–24. (Shifting to plant-based production could allow regrowth sequestering about 332 to 547 Gt CO2, roughly 99% to 163% of the CO2 budget for a 66% chance of 1.5 °C.) nature.com
19. Lindqvist, H., et al. (2014). Alpha-tocopherol and beta-carotene contents of forage species. Grass and Forage Science, 69, 356. doi.org
20. Vasta, V., et al. (2019). Invited review: Plant polyphenols and rumen microbiota responsible for fatty acid biohydrogenation, fiber digestion, and methane emission. Journal of Dairy Science, 102(5), 3781–3804. doi.org
21. Adam, A., et al. (2002). The bioavailability of ferulic acid is governed primarily by the food matrix rather than its metabolism in intestine and liver in rats. Journal of Nutrition, 132(7), 1962–1968. doi.org
22. Serra, V., et al. (2021). Dietary polyphenol supplementation in food producing animals: effects on the quality of derived products. Animals, 11(2), 401. doi.org
23. Kalač, P. (2012). Carotenoids, ergosterol and tocopherols in fresh and preserved herbage and their transfer to bovine milk fat and adipose tissues. Journal of Agrobiology, 29, 1. doi.org
24. Yang, A., Larsen, T. W., and Tume, R. K. (1992). Carotenoid and retinol concentrations in serum, adipose tissue and liver and carotenoid transport in bovine, ovine and caprine species. Australian Journal of Agricultural Research, 43, 1809. doi.org
25. Křížová, L., et al. (2021). Production of bovine equol-enriched milk: a review. Animals, 11(3), 735. (Feed-to-milk isoflavone carry-over rates vary by source and intake and are consistently small fractions of intake.) doi.org. See also Mustonen, E. A., et al. (2009), British Journal of Nutrition, 102(11), 1552–1556. doi.org
26. Carlsen, M. H., et al. (2010). The total antioxidant content of more than 3100 foods, beverages, spices, herbs and supplements used worldwide. Nutrition Journal, 9, 3. (Table 1: plant-based foods median 0.88 and mean 11.57 mmol/100 g; animal-based foods median 0.10 and mean 0.18; the authors note the plant mean is raised by a minority of very high products among plant medicines, spices, and herbs, and that antioxidant content measured in vitro does not necessarily correspond to activity in a target cell.) doi.org
27. Lindeman, R. L. (1942). The trophic-dynamic aspect of ecology. Ecology, 23(4), 399–417. doi.org

Last updated: October 6, 2026 at 6:55 PM UTC