Does ultra-processed food make you eat more?

In a hospital research unit where food was provided and measured, people ate 508 more calories a day on an ultra-processed diet and gained 0.9 kg in two weeks. That is the most reliable result in this area. The same paper shows why the simple reading is wrong. The calorie density of what people actually ate was not matched, and the authors say the 85 percent higher calorie density of the food other than drinks likely played a part.

Is it the processing itself?

The study's own data points to calorie density and eating speed instead. People ate 17 calories a minute faster on the ultra-processed diet. They reported no difference at all in how pleasant or satisfying the food was, which undercuts the too-tasty-to-stop story. The authors say no cause-and-effect link between ultra-processed food and human obesity has been shown yet.

So can a diet high in ultra-processed food be healthy?

The 2025 study is the best answer anyone has. When an ultra-processed diet and a minimally processed diet were both built to follow national healthy eating guidelines, both led to weight loss, and minimal processing won by about one percent of body weight over eight weeks. That suggests what is in the food matters far more than how processed it is, which is useful if cooking from scratch is not realistic for you.

Why do the two trials disagree?

They do not, quite. In 2019 a typical ultra-processed diet led to weight gain. In 2025 an ultra-processed diet that followed healthy eating guidelines led to weight loss. Together they say that what is in the food drives most of the effect, and that ultra-processed is a category broad enough to hold both.

What about the studies linking it to diabetes?

Those studies only watched people. There are 14 of them, covering 692,508 people, with a risk score of 1.24 for the highest intake against the lowest. The direction is consistent, but other factors muddy it badly. People who eat the most ultra-processed food differ in income, education, time and kitchen access, and statistical adjustment does not fully remove that. These studies give length where the trials give control, and neither gives both.

Whole-food2 studies in people cited

Whole food versus ultra-processed

Written by Aaron CuhaReviewed Sep 2026

Also called: NOVA, UPF, Minimally processed

The landmark hospital study found people ate 508 more calories a day on ultra-processed food. Its own data shows the cause was how calorie-dense the food was and how fast people ate it, not the processing itself. A 2025 repeat, where both diets followed healthy eating guidelines, shrank the gap to one percent of body weight.

What it actually is

Not a diet with rules, but a claim about what matters most. The argument is that how heavily food is processed in factories, more than its mix of carbs, fat and protein, explains most of what diet studies find. Everything here depends on the NOVA system for sorting foods. That is a published definition, not something measured, so every study below rests on it.

How its fans say it works

Factory processing makes food that is calorie-dense, soft, quick to eat and built to taste great, and that overrides the signals that tell you that you are full. The strong version says processing itself is the cause. The weak version says processing just goes along with calorie density, eating speed and nutrient content, which are the real causes. Those are different claims, and the best study tells them apart.

What happened when calories were controlled

Best kind of study available: people ate freely

Both groups ate as much as they wanted. A difference here mixes the diet with how full it kept people and how well they stuck to it.

Calories were deliberately not controlled, because how much people ate was the thing being measured. What was matched was the food people were offered. The key finding is that the matching fell short in exactly the way that matters.

Dicken 2025 is the repeat that changes the picture. 55 adults each ate two diets for eight weeks, as much as they wanted, one after the other. Both diets followed the UK's healthy eating guide. One was minimally processed and one was ultra-processed. Both led to weight loss. The minimally processed diet took off 2.06 percent of body weight and the ultra-processed one 1.05 percent. The difference, 1.01 percent, was real but modest, with a likely range of 0.14 to 1.87 percent.
Hall 2019 admitted 20 adults with steady weight to the NIH Clinical Center. Each ate two weeks of ultra-processed food and two weeks of unprocessed food, one after the other, as much as they wanted. Meals were designed to match on calories offered, calorie density, carbs, fat, protein, sugar, salt and fibre. People ate 508 more calories a day on ultra-processed food, a clear difference. They gained 0.9 kg on it and lost 0.9 kg on the unprocessed diet.
The same paper undercuts the simple reading. The calorie density of what people actually ate was not matched: 1.36 against 1.09 calories per gram. The offered meals only matched because the drinks had fibre supplements mixed in. The authors say the roughly 85 percent higher calorie density of the food other than drinks likely added to the extra eating.
Taste was not different either. People reported no real difference in how pleasant, familiar or satisfying the food was, or in how much they felt able to eat. The idea that ultra-processed food is too tasty to stop eating is not backed by what people said in the study most often quoted for it.
Eating speed was different. People ate 17 calories and 7.4 g a minute faster on ultra-processed food, and eating speed tracked the extra eating. So what this study supports is closer to the calorie density and eating speed of the food than to processing itself. The authors write that no cause-and-effect link between ultra-processed food and human obesity has been shown yet.

What happened when people just ate it

The studies that tracked people over years are large and point the same way. A 2025 review pooled 14 long-term studies covering 692,508 people. Those who ate the most ultra-processed food had a risk score of 1.24 for getting diabetes compared with those who ate the least, where above one means higher risk. The score was 1.13 for every 10 percent more of the diet. The results varied a lot from study to study. These studies only watch people, so they carry the usual problem in a strong form. People who eat the most ultra-processed food differ from those who eat the least in income, education, time, kitchen access and much else, and no statistical adjustment fixes that. This evidence still matters because the trials last two and eight weeks, while these studies run for years.

Protein, and whether it muddies the result

Protein is the quiet finding in the hospital study, and it is worth pulling out. The extra 508 calories a day came from carbs (280 more) and fat (230 more), and not at all from protein (2 fewer, no real difference). Protein intake was basically the same on both diets. That fits the protein leverage idea: people keep eating until they hit a protein target, so they eat more total calories when food is low in protein. The authors themselves suggested future studies should match diets more closely for protein and for the calorie density of the food other than drinks.

The measured intakes behind the protein figures are on the protein page.

What reliably moves

Measure	Direction	From
Calories eaten	Up by about 508 calories a day on ultra-processed	The headline result, from a hospital research unit where food was provided and measured. The most reliable number in this section.
Body weight	Up 0.9 kg in two weeks on ultra-processed, down 0.9 kg on unprocessed	Weight change tracked calories eaten closely. In the 2025 repeat, where both diets followed healthy eating guidelines, both groups lost weight.
Body fat	Up 0.4 kg on ultra-processed, down 0.3 kg on unprocessed	From the full paper. The authors put the bigger swings in lean mass down to shifts in body water, and differences in salt intake tracked those lean mass changes.
Eating speed	Faster on ultra-processed	17 calories and 7.4 g more a minute, tracking the extra eating. This, along with calorie density, is what the study's own data supports as the cause.
Reported taste	No difference	No real difference in how pleasant, familiar or satisfying the food was. The most popular explanation for this study's result is not backed by the study.
New cases of type 2 diabetes	Up, in studies that only watched people	A risk score of 1.24 for the highest against the lowest intake, across 14 studies and 692,508 people, and 1.13 for every 10 percent more of the diet. These studies only watch people, and differences in income and education muddy them heavily.

Long term, and serious health events

No trial has run longer than eight weeks, and none has counted heart attacks, strokes or deaths. The long-term studies give the long-term signal but cannot show cause. The most useful thing here is to compare the two trials rather than read either alone. In 2019, an ultra-processed diet led to weight gain over two weeks. In 2025, an ultra-processed diet built to follow national healthy eating guidelines led to weight loss over eight weeks, and the edge for minimal processing shrank to about one percent of body weight. Those results do not clash. Together they say that what is in ultra-processed food, not the processing itself, drives most of the effect. They also say a guideline-following ultra-processed diet behaves very differently from a typical one. That is a more useful conclusion than the headline either study made.

Sources

Study in people2019
Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake
Cell Metabolism
20 adults lived on an NIH hospital ward and ate ultra-processed and unprocessed diets, two weeks each. They ate 508 more calories a day on ultra-processed food, from carbs and fat, not protein, and gained 0.9 kg, against losing 0.9 kg on unprocessed. The ultra-processed food packed more calories per bite and was eaten faster. People liked both diets equally.
20 adults, NIH ward, two weeks per diet in crossover. Energy intake 508 kcal/day greater on ultra-processed (p = 0.0001), from carbohydrate and fat but not protein. Weight plus 0.9 kg on ultra-processed and minus 0.9 kg on unprocessed. Consumed energy density was not matched (1.36 against 1.09 kcal/g) and eating rate was faster; palatability did not differ.
Study in people2025
Ultraprocessed or minimally processed diets following healthy dietary guidelines on weight and cardiometabolic health: a randomized, crossover trial
Nature Medicine
55 adults ate as much as they wanted on two diets for eight weeks each, both following UK healthy eating advice. Both led to weight loss: 2.06 percent on minimally processed food and 1.05 percent on ultra-processed, a gap of 1.01 percent. The authors listed many outside ties, on both sides of this debate.
55 adults, two eight-week ad libitum diets both following the UK Eatwell Guide. Both produced weight loss: minimally processed minus 2.06 percent and ultra-processed minus 1.05 percent, difference 1.01 percent (p = 0.024, Cohen's d minus 0.48). Extensive competing interests declared on both sides of the topic.
Review of studies2019
Ultra-processed foods: what they are and how to identify them
Public Health Nutrition
This is the NOVA system, the definition of ultra-processed food that every study here relies on. It is a set of labels, not something measured. So any flaw in it carries into all of this research.
The NOVA classification reference on which every study in this block depends. A published definition rather than a measured quantity, which makes it an assumption carried into all of this evidence.
Review of studies2025
Ultra-processed food consumption and risk of type 2 diabetes: a systematic review and updated meta-analysis of prospective cohort studies
Metabolism
A review pooled 14 studies that followed 692,508 people over time. Those eating the most ultra-processed food had more type 2 diabetes than those eating the least, and each 10 percent rise in intake came with more risk. These studies track people and cannot prove cause, and income and education muddy the results.
14 prospective cohorts, 692,508 participants. Highest against lowest ultra-processed consumption HR 1.24 (1.14 to 1.34), I squared 69 percent. Per 10 percent increase in intake HR 1.13 (1.08 to 1.18). Observational, with strong socioeconomic confounding.

What people report

These are things people say about their own eating. Nobody checked them, so they are not evidence. They are here because they tell you what to expect and what to watch for, which the studies do not. They cannot tell you whether anything works.

Less snacking and less frequent hunger after switching to home-cooked food, reported very widely. That fits the findings on calorie density and eating speed.
A lot more time spent cooking and shopping, the most common reason people stop.
Confusion about what counts. That is fair when wholemeal bread and a fizzy drink share a category.
Higher grocery bills reported often, though the studies say nothing about cost.
Eating more slowly, described as the change people notice most. That happens to be the thing the hospital study measured as different.

Sources: General nutrition groups and the growing pile of popular books on ultra-processed food, some written by authors with money riding on the topic. That cuts both ways. The conflict-of-interest statements in the 2025 study include book royalties on ultra-processed food, broadcasting fees on the subject, drug company jobs and shares, and clinic ownership. Otherwise unchecked self-report.

Who this is wrong for

Nobody, since this is not a diet plan, but one warning is worth stating. Treating a NOVA category as a health verdict will mislead you both ways. Supermarket wholemeal bread and a sugary drink are both classed as ultra-processed.
Anyone for whom cooking from scratch is not realistic because of time, money or kitchen access. The 2025 study shows a guideline-following ultra-processed diet still led to weight loss. That is truly useful for people whose problem is not willpower.
Anyone using this idea to avoid fortified foods. Foods with added B12 count as processed, and they are the practical way many people get enough, especially on plant-based diets.
Anyone reading the long-term study data as cause and effect. Income and education muddy ultra-processed food data more than almost anything else in nutrition research.

Questions

Does ultra-processed food make you eat more?: In a hospital research unit where food was provided and measured, people ate 508 more calories a day on an ultra-processed diet and gained 0.9 kg in two weeks. That is the most reliable result in this area. The same paper shows why the simple reading is wrong. The calorie density of what people actually ate was not matched, and the authors say the 85 percent higher calorie density of the food other than drinks likely played a part.
Is it the processing itself?: The study's own data points to calorie density and eating speed instead. People ate 17 calories a minute faster on the ultra-processed diet. They reported no difference at all in how pleasant or satisfying the food was, which undercuts the too-tasty-to-stop story. The authors say no cause-and-effect link between ultra-processed food and human obesity has been shown yet.
So can a diet high in ultra-processed food be healthy?: The 2025 study is the best answer anyone has. When an ultra-processed diet and a minimally processed diet were both built to follow national healthy eating guidelines, both led to weight loss, and minimal processing won by about one percent of body weight over eight weeks. That suggests what is in the food matters far more than how processed it is, which is useful if cooking from scratch is not realistic for you.
Why do the two trials disagree?: They do not, quite. In 2019 a typical ultra-processed diet led to weight gain. In 2025 an ultra-processed diet that followed healthy eating guidelines led to weight loss. Together they say that what is in the food drives most of the effect, and that ultra-processed is a category broad enough to hold both.
What about the studies linking it to diabetes?: Those studies only watched people. There are 14 of them, covering 692,508 people, with a risk score of 1.24 for the highest intake against the lowest. The direction is consistent, but other factors muddy it badly. People who eat the most ultra-processed food differ in income, education, time and kitchen access, and statistical adjustment does not fully remove that. These studies give length where the trials give control, and neither gives both.