You can lose weight with the help of artificial intelligence

Artificial intelligence can't melt away pounds, speed up metabolism, or undo the excess food we've eaten. But it can do something that makes most weight loss attempts fail: turn the general decision "I need to eat less" into a concrete, everyday system that's tailored to real life.

When asked if artificial intelligence can help you lose weight, the most honest answer is: it can, but not because AI knows some secret that nutritionists are hiding. Its value is not in a miracle diet, but in planning, monitoring, recognizing patterns and providing support at the very moments when discipline usually falters.

In other words, not a skinny algorithm. A skinny human who, thanks to a well-used algorithm, makes it easier to make good enough decisions – for long enough.

Kilograms are not just a mathematical equation

To lose weight, you need to consume less energy than your body burns over time. But that doesn't mean that losing weight is simply a matter of adding calories. Weight is influenced by genetics, hormones, sleep, stress, medications, health conditions, food availability, habits, emotions, and environment. The World Health Organization now describes obesity as a chronic, relapsing, and complex condition, not just a lack of character or self-control.

This is where AI comes in. It can’t change biology, but it can reduce everyday chaos: plan meals in advance, create a shopping list, offer a substitute for a higher-calorie food, warn against repetitive evening snacking, or show that the problem may not be lunch, but wine, juices, and “just a few bites” during the evening.

The evidence is neither sensational nor insignificant. Most research so far has focused not on simple chatbots, but on structured digital programs that combine weight and nutrition tracking, education, feedback, goal setting, and occasional expert support. That distinction is important: talking to a generic AI tool is not the same as a clinically proven weight management program.

A systematic review published in 2025, which included 16 studies of automated programs, concluded that none of the interventions analyzed led to an average reduction of five percent from baseline weight. Regular encouraging messages, education based on the principles of cognitive behavioral therapy, and continuous self-monitoring showed the most potential, especially when combined with human contact.

However, the results of individual, well-designed programs can be significantly better. In a randomized trial of 164 obese adults, a structured digital app led to an average weight loss of 5,29 percent over six months, compared with 1,76 percent in the group receiving usual care. The program was not just an app: it included meal plans, tracking, education, goals, feedback, and contact with a dietitian.

Another large clinical trial, with 540 participants, showed that an automated online program could lead to an average weight loss of 3,6 kilograms over the first three months. Participants who then received active support to maintain their results regained less weight over the next two years than those who received only monthly newsletters. More days of follow-up and greater program engagement were associated with less weight regain.

The message of the research is therefore not that “AI is getting skinny,” but that a well-organized digital system can help people be more consistent. And consistency is usually more important than perfection.

Where AI can be particularly useful

He can make a plan that resembles your life.

Classic menus often assume that a person eats breakfast at eight, lunch at two, cooks every day, loves all foods, and has no business dinners, children, allergies, or stress. Such a plan may be nutritionally correct, but practically unusable.

AI can take into account work hours, budget, groceries available in Montenegro, allergies, family meals, time of first meal, days when cooking and days when eating out. It can offer multiple variations of the same meal and come up with a plan that doesn't require a person to become a completely different person overnight.

The best plan is not the one that looks the healthiest on paper, but the one that can be repeated even when the day is tough.

It can reveal where the plan is actually failing.

People often misjudge what is holding them back from their goals. Someone believes they eat too much at lunch, but their diary shows that the biggest surplus occurs after dinner. Another eats very little during the workday, then makes up for their hunger in the evening. A third eats disciplined from Monday to Friday, but over the weekend they cancel out their entire weekly deficit.

When fed regular food, drink, hunger, sleep and body weight data over several weeks, AI can summarize patterns that are difficult to see from a single day. For example, it can show that snacking is most common after poor sleep, that portions increase after skipping meals, or that alcohol is regularly accompanied by additional food.

This is where AI works best as a mirror, not as a judge.

It can reduce the number of decisions we make when hungry.

Weight loss often fails not because we don't know that vegetables and fish are better choices than pastries and fried foods. It fails because at seven in the evening, tired and hungry, we are just starting to decide what to eat.

AI can pre-prepare three quick dinners from ingredients we already have, suggest better choices from a restaurant menu, come up with a substitution when we have a sweet tooth, or make a plan for situations like travel, celebrations, and business lunches. This way, we don't rely solely on willpower, which is especially unreliable when we're exhausted.

A bigger challenge than losing weight is often keeping it off. Digital programs that provide ongoing monitoring, feedback, and occasional reminders show better weight maintenance results than programs that end abruptly.

AI can therefore be useful even after a diet: to recognize gradual increases in portions, reduction in movement, or a return to old habits before a few hundred grams turn into several kilograms.

A photo of a meal is not a laboratory analysis

One of the most appealing options is to take a photo of a plate and ask: “How many calories does this have?” Such an estimate can be useful as a framework, but not as a precise measurement.

AI can't tell from a photo how much oil was used, what's under the sauce, what the actual weight of a portion is, or whether the ingredients were fried, baked, or boiled. A study published in 2026, which tested 40 visual-linguistic models, found that professional nutritionists were significantly more accurate than all the systems tested, especially when estimating protein.

Another study showed that results improved when the user, in addition to the photo, provided the name of the food, quantity, and preparation method. The most common errors were portion estimates, hidden ingredients, unclear photos, and made-up or misidentified ingredients.

That's why a photograph is good as a diary and reminder, but not as an infallible calorimeter. When greater precision is important, you should use a product label, a scale, or a verified nutritional database.

Large language models can sound convincing even when they are wrong. A systematic review of their use in obesity treatment found that they can be useful for education, motivational interviewing, and personalizing recommendations, but also produce inconsistent, sometimes inaccurate, or biased responses, especially in complex medical situations.

But that's why you don't need to:

In case of health problems, a weight loss plan should be coordinated with a doctor or registered dietitian. The US National Institute of Diabetes and Digestive and Kidney Diseases specifically emphasizes that a successful program should have realistic goals, an individually tailored plan, monitoring of habits and weight, regular feedback, and a strategy for maintaining results.

Privacy is also important. Data on weight, diseases, therapy, meals and photos represent sensitive personal information. The World Health Organization warns that health AI systems must protect the autonomy, privacy, security and accountability of users. Therefore, publicly available chatbots should not include names, personal data documentation or more medical details than are really necessary.

How to talk to AI so that the response is truly useful

A bad request is: "Make me a diet to lose weight quickly."

A much better request could look like this:

I want to lose weight gradually and sustainably. I am ___ years old, ___ centimeters tall, ___ kilograms, and my physical activity level is ___. I work from ___ to ___, usually eat my first meal at ___, sleep approximately ___ hours, do not eat ___, and am allergic to ___. Of the health conditions and therapies, ___ is relevant. My biggest problem is evening snacking/skipping meals/weekend overeating.

Do not make diagnoses, prescribe medications, or impose extreme restrictions. State any assumptions you are making and mark any estimates that are not accurate. Suggest a realistic seven-day meal plan from available foods, with approximate serving sizes, simple substitutions, a shopping list, and a plan for when you eat out. At the end of each week, analyze your average weight, hunger, energy, sleep, and adherence to the plan and suggest up to two changes for the following week.

It is also useful to ask the AI ​​to explain why it is suggesting something, to offer several options instead of a single prohibition, and to clearly separate known facts from assessments.

It can - if we use artificial intelligence as an organizer, reminder, habit analyzer, and decision-making aid. It can't - if we expect it to be an infallible nutritionist, doctor, moral authority, or a substitute for our own behavior.

Its greatest value is not in putting together the "perfect diet." Perfect diets tend to be short-lived. Its value is in making an imperfect but sensible plan easier to follow.

AI can show the way, notice when we've strayed from it, and suggest how to get back on track. But we still have to take the steps ourselves.

Article prepared with the support of AI tools.

Download the app and follow the news