For decades, the conventional wisdom was as simple as it seemed intuitive: eating fat makes you fat, and fat accumulation causes fatty liver disease. So when researchers began documenting that a diet composed of up to 80 percent fat could actually reduce liver fat in some patients, the hepatology community took notice, and largely couldn’t believe what it was seeing.
The ketogenic diet, which severely restricts carbohydrates to between 20 and 50 grams per day while dramatically elevating fat intake, has become one of the most polarizing nutritional interventions in modern medicine. But while social media influencers and diet book authors tout keto as a metabolic cure-all, the actual clinical evidence tells a considerably more complicated story: one involving real therapeutic promise, significant biological risks, and a liver caught at the center of a remarkable physiological tug-of-war.
What Happens Inside Your Liver When You Go Keto
To understand the controversy, you first need to understand what the liver does when the body’s carbohydrate supply is cut off. Under normal dietary conditions, the liver converts excess glucose into triglycerides through a process called de novo lipogenesis: essentially manufacturing fat from sugar. When dietary carbohydrates drop precipitously, this process slows dramatically. Insulin levels fall, and the pancreas signals fat cells to begin releasing their stored triglycerides into the bloodstream as free fatty acids.
The liver intercepts much of this flood of liberated fat. It begins oxidizing these fatty acids at elevated rates, and crucially, converts a portion of them into ketone bodies (acetoacetate, beta-hydroxybutyrate, and acetone), which become the primary fuel source for the brain and other organs in the absence of glucose. This metabolic state is ketosis, and it’s the intended goal of the ketogenic diet.
Here is where the biology becomes genuinely fascinating and medically contested. The liver is simultaneously processing far more fat than usual and suppressing its own internal fat production. For people with nonalcoholic fatty liver disease (NAFLD), a condition affecting approximately 25 percent of the global adult population, or roughly two billion people, a major driver of liver fat accumulation is not dietary fat itself but de novo lipogenesis fueled by excess carbohydrates and sugar. In a widely cited 2005 isotope study in the Journal of Clinical Investigation, about 59 percent of liver triglyceride in NAFLD patients came from free fatty acids released by fat tissue, about 26 percent from de novo lipogenesis (a pathway that keto directly throttles) and about 15 percent from dietary fat.
What the Clinical Trials Actually Show
The clinical evidence on keto and liver fat is small in volume but suggestive, at least in the short term.
A striking 2020 study published in PNAS by researchers at Yale and the University of Helsinki put 10 overweight or obese people with nonalcoholic fatty liver disease on a ketogenic diet for just six days. Using MRI-based fat quantification, the researchers found that liver fat fell by 31 percent despite only a small drop in body weight, while the liver’s insulin resistance improved sharply.
Other small controlled trials have found that very-low-carbohydrate ketogenic diets cut liver fat faster than low-fat or low-calorie diets over days to weeks. One comparison, however, found no difference between the two diets after about 11 weeks once weight loss was equal, a reminder that much of the benefit may reflect calorie reduction and weight loss as well as ketosis.
Reviews of the small trials in this area broadly suggest that ketogenic and very-low-carbohydrate diets can reduce liver fat in the short term, but the studies are small and short, and the advantage over other diets with similar weight loss is not firmly established.
The mechanism driving this improvement is reasonably well understood. By suppressing insulin and dramatically reducing de novo lipogenesis, the ketogenic diet removes the primary metabolic input that causes the liver to accumulate fat in the typical NAFLD patient. Simultaneously, the upregulation of hepatic fat oxidation accelerates clearance of existing fat stores. The liver, in effect, stops filling up and starts emptying out.
Only two drugs have been approved by the FDA for NASH, the more aggressive inflammatory form of the disease: resmetirom (Rezdiffra) in 2024 and semaglutide (Wegovy) in 2025, both for patients with moderate-to-advanced fibrosis, so dietary interventions remain not just interesting academic questions but urgent clinical priorities. NASH is now among the leading causes of liver transplantation in the United States. Against this backdrop, a dietary intervention showing reductions in liver fat of around 30 percent within days, as in the 2020 PNAS study, is notable, even if its long-term value is unproven.
The Long-Term Problem: When Healing the Liver Gets Complicated
The short-term data are compelling. The long-term data are almost nonexistent, and what exists raises real questions.
The first problem is adherence. Ketogenic diets are metabolically demanding and socially isolating. Drop-out rates in clinical trials can be substantial, even in closely supervised settings where participants receive regular support and guidance, and many people in real-world populations abandon strict keto within months. Whatever the liver benefits of sustained ketosis, they require sustained ketosis.
More troubling are emerging data on lipid profiles. While keto reliably reduces triglycerides, sometimes dramatically, it frequently raises LDL cholesterol, and specifically the apolipoprotein B (apoB) fraction, which is considered a more reliable marker of cardiovascular risk than total LDL. Reports have described a subset of lean individuals on carbohydrate-restricted diets who developed profound hypercholesterolemia, in some cases LDL levels above 300 mg/dL, a pattern researchers dubbed “lean mass hyper-responders.” Whether this translates into elevated cardiovascular risk over years or decades remains unresolved, but it is not a question clinicians can ignore.
The diet’s effect on the liver itself may also become more complicated over extended periods. Animal studies (which must be interpreted cautiously but often prefigure human outcomes) have produced conflicting results. Some mouse models show reversal of NAFLD on ketogenic diets; others, particularly those using very high-fat compositions over longer periods, show paradoxical worsening of liver inflammation and the development of hepatocellular carcinoma markers. In some of these studies, mice on a ketogenic diet showed early resolution of liver fat followed by a different pattern of liver damage, with oxidative stress and inflammatory markers rather than lipid accumulation. Mouse results do not necessarily carry over to people.
Whether the liver can sustain years of heightened fat oxidation is not known, and clinicians are cautious about extrapolating short-term human data to long-term recommendations.
The NAFLD Spectrum: Keto Doesn’t Work the Same Way for Everyone
One of the most important nuances in the clinical literature is that fatty liver disease is not a monolithic condition, and keto does not affect all NAFLD patients equivalently.
NAFLD exists on a spectrum from simple steatosis (fat accumulation without significant inflammation or damage) to NASH, which involves inflammation and hepatocyte injury, to advanced fibrosis and cirrhosis. Patients at different points on this spectrum respond differently to dietary interventions.
For patients with simple steatosis driven primarily by excess carbohydrate intake and insulin resistance (the most common presentation in Western populations), the ketogenic approach appears most effective. This phenotype responds well precisely because keto targets the dominant pathological pathway.
However, for patients who already have significant fibrosis, or whose NAFLD has a meaningful genetic component (variants in the PNPLA3 gene, for example, dramatically increase NAFLD risk independent of metabolic status), the situation is less clear. Most of the encouraging trial data also come from patients without advanced fibrosis, so how well the results carry over to more advanced disease is uncertain. Similarly, patients with coexisting alcohol-related liver injury or viral hepatitis require entirely different management considerations.
There is also a meaningful question about composition. Not all ketogenic diets are metabolically equivalent. A keto diet high in saturated fats from red meat and butter may have different effects on liver inflammation markers than one built around olive oil, avocados, and fatty fish, despite both inducing ketosis, although direct comparisons in people are limited.
The Gut-Liver Axis: An Overlooked Dimension
One emerging and underreported dimension of keto’s liver effects involves the gut microbiome: the complex ecosystem of bacteria, fungi, and other microorganisms that populate the gastrointestinal tract and communicate directly with the liver via the portal venous system.
The liver receives approximately 70 percent of its blood supply directly from the gut. Bacteria in the gut produce a wide range of metabolites, including lipopolysaccharides (LPS), short-chain fatty acids, bile acids, and ethanol, that flow directly to the liver and profoundly influence its inflammatory state. Dysbiosis, or disruption of normal gut microbial composition, is increasingly recognized as a significant driver of NAFLD progression.
Ketogenic diets alter the gut microbiome substantially. The dramatic reduction in dietary fiber (a consequence of eliminating most fruits, grains, and legumes) reduces the substrates available for beneficial fiber-fermenting bacteria such as Bifidobacterium and Lactobacillus species. Multiple studies have documented reduced microbial diversity on ketogenic diets, and some have shown increases in LPS-producing gram-negative bacteria, which could theoretically worsen hepatic inflammation via the gut-liver axis even as the diet reduces liver fat through metabolic mechanisms.
This creates a genuinely paradoxical possibility: keto might simultaneously be reducing liver fat while increasing a distinct inflammatory input that could worsen liver disease over time. The clinical significance of this possibility is not yet established, but it is being actively studied.
Where the Science Is Heading, and What Patients Should Know
The honest scientific picture of keto and liver health today looks something like this: there is compelling short-term clinical trial evidence that ketogenic diets reduce liver fat faster and more substantially than standard low-fat diets in patients with NAFLD driven by metabolic dysfunction. This is not a trivial finding in a disease landscape where approved drug treatments are only just emerging and remain limited to patients with more advanced disease.
But the long-term data simply do not exist in sufficient quantity or quality to make definitive recommendations. Larger, longer randomized trials with liver biopsy endpoints are needed to fill this gap, and few have yet reported results.
In the meantime, several conclusions seem reasonably defensible. First, a short-term ketogenic diet (in the range of three to six months) under medical supervision may be a reasonable option for reducing liver fat in appropriate NAFLD patients, particularly those with concurrent insulin resistance, obesity, and metabolic syndrome. Second, the quality of dietary fat consumed on a ketogenic diet matters significantly; a version built around unsaturated fats such as olive oil is likely a more sensible choice than one high in saturated fat, though the evidence is limited. Third, the diet is not suitable for everyone: patients with pre-existing renal disease, a history of pancreatitis, or significant cardiovascular disease require careful clinical evaluation before attempting it.
Perhaps most importantly, the story of keto and the liver is a useful corrective to dietary fundamentalism in both directions. The anti-fat camp that dismissed keto as inherently hepatotoxic was wrong, at least in the short term and in specific patient populations. But the keto evangelists who claim the diet is universally healing and without risk are equally misleading. The liver, as it turns out, doesn’t care about dietary ideology. It responds to metabolic signals, and those signals are more nuanced than any single nutritional framework can capture.
The most useful thing the emerging science on keto and NAFLD may ultimately teach medicine is not that high-fat diets are good or bad, but that the relationship between nutrition and organ health is mechanistically specific, and that treating it as a simple morality play about “clean eating” or “dangerous fats” has cost patients time and clinicians insight that a more rigorous, molecular approach might have delivered years earlier.