“I thought it was instagram. That I was looking at impossible bodies, strange diets. It never occurred to me that I was asking ChatGPT how to leave half a cup of yogurt on the plate so that it looked like I had had breakfast,” says Andrea.
Martina’s mother—her name was changed to preserve her identity—had the password to her 15-year-old daughter’s cell phone “for security.” I had never opened it. But, little by little, various behaviors began to worry her.
Martina started the summer saying she wanted to “get organized.” That I was going to eat better, train more. Nothing that would attract too much attention. But as the weeks went by, she stopped having breakfast with her family and avoided going out with friends to fast food places, which she previously “enjoyed a lot.” He said that “that vibe no longer suited him.” He used to go to his room with his cell phone after each meal or say that he had already had dinner when the plate was still almost untouched.
When Her mother looked at the phone screen out of the corner of her eye, and the teenager responded that she was “researching” about healthy eating. Until one morning he discovered her sitting at the kitchen counter, carefully reading the nutritional information labels of various foods. That night, he waited for her to fall asleep and decided to look at her cell phone.
ChatGPT was open and accumulated several questions from the teenager:
—If I ate two pieces of toast and I feel guilty, how do I compensate?
—How many calories does a salad have without dressing?
—How do I make sure my parents don’t notice that I’m eating less?
— What are the exercises to burn more fat?
“I spoke to the chat as if it were a friend and asked a barrage of questions related to food, body image and physical activity”, describes Andrea and continues: “When the questions were very borderline, like when she asked how to vomit quickly, the AI gave her an alert to ask for help. But she learned to circle the questions to get some kind of answer.”
Martina detailed everything she had eaten that day and asked the chat to tell her how many calories those foods had. “I sent her her weight, her height. I asked her if she was fat. It was like having someone available 24 hours a day who would answer exactly what she wanted to hear”adds his mother.
Thanks to that discovery they were able to ask for professional help. “Three months ago he was diagnosed anorexia and today he is in treatment, taking baby steps little by little,” he says.
In the offices, Psychiatrists and nutritionists specialized in eating disorders are frequently seeing cases like Martina’s: adolescents—generally women between 13 and 17 years old—who use tools of artificial intelligence to validate food restrictions, calculate calories and reinforce control behaviors typical of eating disorders (ED).
“Since last year I have seen it more and more,” says Alejandra Freire, nutritionist at the Surgery Service of the Hospital de Clínicas and certified specialist in eating disorders by the International Association of Eating Disorders Professionals Foundation.
“One day in the office, I asked a patient to show me what she was asking ChatGPT. When we opened it together, a world opened up,” she says. “Many teenagers perceive AI chat as an expert, almost unquestionable voice,” warns Juana Poulisis, a psychiatrist with extensive experience in eating disorders
The queries are varied: requests for “1,000-calorie diets,” personalized plans, questions about how to make their parents believe they had lunch. Some patients even send their own photos to ask opinions about their body.
“The key is how they ask,” explains Freire. “If you directly consult something risky, the system alerts. But teenagers know how to use the prompt. They circle the question, pose a hypothetical case, talk about ‘a friend’, formulate the query in the positive. “They have an ability that we adults don’t have.”
Juana Poulisis, a psychiatrist with extensive experience in the subject and fellow of the World Academy of Eating Disorders, describes it like this: “Artificial intelligence has become the new oracle of our time. Like the Oracle of Delphi, it has enormous symbolic authority. Many adolescents perceive it as an expert, almost unquestionable voice”.
Risk is not only what responds
Why is this trend worrying? For Poulisis, the problem is not only the specific content of the responses. “The greatest risk lies not only in what the AI says, but in how it interacts: the progressive personalization, the tone, the frequency. It speaks to you, with your data. It is available 24/7. You can be at three in the morning reinforcing ruminative thoughts about your body.”
This constant availability, without waiting lists or schedules, can become an amplifier of symptoms. “Feedback loops are generated. The damage is not always immediate or obvious. “They arise from microinteractions that at first seem harmless and that, over time, create problematic patterns,” points out.
In addition, he warns, AI facilitates continuous rumination: “By always being accessible, it becomes a tool for repetitive and intrusive thoughts about weight, food, and physical appearance.”
Among the risks, it lists: health advice without a medical context, normalization of restrictive behaviors, body comparison, reproduction of models of idealization of thinness – what is known as thinspiration -, link between weight and personal worth and moralization of foods.
Freire adds another element: “Many times, when the chat tells you what not to do—not vomit, not fast—it ends up listing behaviors that the adolescent had not even considered. That information can give ideas.”
Unlike Google, they explain, AI does not just summarize information: it organizes and personalizes it. “We put together a diet adjusted to your weight, your height, your activity. That gives a feeling of precision and support that a traditional search engine does not have,” says Freire.
Specialists state that the AI chat ends up listing behaviors that the teenager had not even considered and that can be dangerous
The phenomenon also has another spice: the chats operate anonymously. History can be deleted. There are no witnesses.
Last November, LA NACION published the case of a father who reported that ChatGPT had accentuated his 14-year-old daughter’s eating disorder and launched a petition on Change.org to request greater warnings and controls on these platforms.
“Parents are very attentive to Instagram: what accounts their daughters follow, what content they watch. But they have no idea of the power that AI has,” warns Freire and adds: “For many children it is like their best friend: always available, free and non-judgmental.”
Can there be something positive?
The specialists avoid an exclusively alarmist view and also point out a possible window of opportunity. “There are adolescents who are developing an eating disorder and first consult the AI about their symptoms,” says Poulisis. “For example: ‘This happens to me, I binge eat twice a week, can it be a disorder?’ That question can be the beginning of awareness and a request for help. It can function as a bridge to professional care.”
Freire agrees: “Just as they seek help to sustain the behavior, when they begin to feel that the situation dominates them, they also ask what is happening to them. Sometimes a diagnostic approach comes along there.”
The challenge, Poulisis maintains, is that this potential is not left to chance. “We need clinical doctors, health agents, to participate in the design of these technologies. Today AI is not a safe space: they are generalist systems, not intervened. Until recently we thought about EDs from an individual perspective. With the emergence of AI we have to think in terms of a system.” “We need clinical doctors, health agents, to participate in the design of these technologies,” says Alejandra Freire, nutritionist at the Surgery Service of the Hospital de Clínicas.
Open the conversation
“I thought that the danger was in the networks, in the edited photos, in the influencers,” says Andrea. “I never imagined that the risk could be in a chat window that seems harmless. Today I understand that we have to talk about this.”
The specialists agree: as well as promoting digital literacy in social networksnow it is necessary to talk about the critical use of artificial intelligence. Ask what is being consulted, how it is interpreted, remember that no automatic response replaces professional evaluation.
Martina is undergoing treatment. There are good days and more difficult days. At home they talked about food again without it being a battlefield. But Andrea can’t stop thinking about those early mornings in front of the screen.
“What strikes me the most is that she didn’t feel like she was doing something serious. She felt like she was learning. That’s why I think we also have to start asking them what they are asking the artificial intelligence. Because sometimes it’s not that they don’t want to talk. It’s that they are already talking… just not with us,” the mother concludes.