Why Nolis Doesn't Give Medical or Psychological Advice

⚠️ Editorial notice

This article is informational content. It does not constitute medical, psychological, legal advice, or a diagnosis. What you read here is an informed opinion, grounded in the studies and sources cited, reflecting the knowledge available at the time of writing.

If you notice concerning signs in a minor in your care — sudden behavioural changes, withdrawal, self-harm, suicidal ideation, emotional dependency on a digital tool — always consult a qualified professional, whether a mental-health specialist or your paediatrician. In an emergency:

• 988 — Suicide & Crisis Lifeline (US, 24 h)
• 911 — Emergency services (US / Canada)
• 116 123 — Samaritans (UK & Ireland · 24 h · free)
• 999 — Emergency services (UK)

HolaNolis detects warning signs and alerts the responsible caregivers. It does not diagnose and does not replace a professional.

Over the past two years, headlines about artificial intelligence and teenagers have become increasingly alarming. Chatbots underestimating what teenagers need to eat: for simulated profiles of teens with overweight or obesity, five chatbots calculated about 695 kcal a day less than a dietitian did (Frontiers in Nutrition, 2026 — AI diet plans underestimate nutrient intake in adolescents). Virtual assistants suggesting stopping psychiatric medication. Apps that "diagnose" anxiety or depression based on a five-question survey. And most seriously: young people who follow this advice to the letter because they trust AI more than any adult.

When we designed HolaNolis, we made a decision that might seem like a limitation to some: Nolis is designed not to give medical or psychological advice or diagnose: it listens and points to trusted adults and support resources. It's not a bug, not a gap in functionality. It's an ethical, regulatory, and safety decision that we consider one of our greatest strengths.

Why is it dangerous for an AI to give health advice to teenagers?

To understand why this decision matters so much, let's look at what's happening out there.

Blind trust in AI

The most recent studies show an alarming trend: teenagers trust AI responses more than web search results, and in many cases more than their own parents or teachers. There's a psychological explanation:

• AI responds immediately, without judging, without asking "why do you want to know that?"
• AI sounds confident. It doesn't say "I don't know" or "it depends." It generates responses with an authoritative tone that adults rarely use.
• AI is private. A teenager can ask about symptoms of depression without fear of anyone finding out.
• AI is always available. At 3am, when anxiety peaks, there's no other person to turn to.

The real cases

These aren't hypotheticals. They're documented facts:

• Dangerous diets: general-purpose chatbots generating extreme "meal plans" for teenagers — in one test, ChatGPT gave a 13-year-old persona a plan of 800, 500 and 300 calories a day (CCDH, 2025) — without considering that a growing teenager needs significantly more. In a 2026 study with simulated teen profiles, five chatbots calculated about 695 kcal a day less than a dietitian did (Frontiers in Nutrition, 2026).
• Medication tapering: AIs suggesting "natural alternatives" to prescribed psychiatric medication, with potentially fatal consequences.
• Erroneous self-diagnosis: minors convincing themselves they have disorders based on chatbot interactions, either delaying consultation with a professional or developing additional anxiety over a diagnosis that doesn't exist.
• Serious incidents involving companion chatbots: families in the US have sued chatbot companies alleging that the chatbots contributed to their teenagers' suicides, and testified before the US Senate in 2025 (AP, 2025; Google–Character.AI settlement CNN), including allegations that the chatbot failed to detect clear signals of suicidal ideation or, worse, reinforced destructive narratives.

Why it's dangerous

The root of the problem is threefold:

1. AI is not qualified. It has no medical training, cannot examine the patient, doesn't know their history, and cannot run tests. Its "knowledge" is a statistical pattern, not clinical understanding.
2. AI has no legal accountability. If a doctor gives bad advice, there are accountability mechanisms. If an AI recommends a dangerous diet, no one is responsible.
3. Teenagers don't filter. An adult can (though not always) cross-check information. A 13-year-old at 2am searching "how to stop feeling sad" doesn't have that filtering capacity.

HolaNolis's position: companion, NOT advisor

HolaNolis is built on a principle we repeat internally like a mantra: detect, alert, redirect. Never diagnose, prescribe, or advise.

This translates into very specific behavior from Nolis:

What Nolis DOES do

• Reflects feelings: "It sounds like you're feeling frustrated about that situation" or "It seems like today was a tough day." Nolis acts as an emotional mirror that helps the minor identify and name what they're feeling, without interpreting or evaluating.
• Validates without judging: "It makes sense to feel that way when things like this happen" or "A lot of people would feel the same in your situation." Emotional validation is deeply therapeutic and doesn't require being a health professional.
• Encourages talking to trusted adults: "Have you been able to talk to anyone about this? Sometimes telling someone you trust helps a lot" or "Is there an adult you feel comfortable talking to about this?"
• Provides professional helplines: when the situation calls for it, Nolis offers crisis phone numbers and localized resources (Samaritans, Crisis Text Line, and equivalent services in each country).
• Activates crisis alerts: if it detects signs of serious risk (self-harm, suicidal ideation, abuse), it alerts the family immediately by email and, if enabled, by push notification. This happens automatically through the four-layer safety pipeline — though no detection system is infallible: it can get things wrong or miss a signal.

What Nolis is designed NOT to do

• Does not diagnose: it is designed not to say "it sounds like you have anxiety" or "that sounds like depression." It doesn't have the qualifications or information to do so.
• Does not prescribe: it is designed not to suggest medications, supplements, doses, or "natural remedies."
• Does not recommend diets: it is designed not to generate meal plans, count calories, or comment on weight or body image.
• Does not give therapeutic advice: it is designed not to say "you should try this relaxation technique for your anxiety" or "what you need is to talk more about your trauma."
• Does not substitute a professional: if the minor asks for mental health help, Nolis is designed to redirect to qualified people.

Why this makes HolaNolis MORE valuable, not less

It might seem counterintuitive: is a chatbot that doesn't give advice better than one that does? Absolutely, and for several reasons.

1. Safety as a value proposition

For a parent, knowing that the AI their 13-year-old talks to is designed not to suggest a diet, a diagnosis, or stopping medication is invaluable. It's the difference between leaving your child at a pool with a lifeguard or without one.

2. The bridge effect

Nolis doesn't try to solve health problems; it tries to build a bridge toward those who can. A teenager who talks to Nolis about their sadness and receives the suggestion to speak with an adult is more likely to do so than one who never verbalized that sadness at all. Nolis is the first step, not the last.

3. Listening without consequences

Sometimes a teenager simply needs to say out loud how they're feeling without anyone going into "solution mode." Nolis offers exactly that: a space to express emotions without receiving a diagnosis, a judgment, or an unsolicited action plan.

4. Early detection

Paradoxically, a system that doesn't give advice but detects and alerts may be more effective than one that tries to treat. Because early detection by an AI system, combined with intervention from a human professional, is an extraordinarily powerful combination. To learn how to identify these signals as a parent, read 5 signs your child needs emotional support.

The regulatory advantage under the EU AI Act

The decision not to give advice isn't just ethical; it's also strategically sound from a regulatory standpoint.

The European AI Regulation (EU AI Act), which entered into force in 2024 and has applied in stages since February 2025, is usually explained in four risk levels:

• Unacceptable risk: prohibited (subliminal manipulation, social scoring…)
• High risk: strict requirements and conformity assessment
• Transparency obligations (often called "limited risk", a label the Regulation itself doesn't use)
• Minimal risk: free use

An AI that gave medical or psychological advice to minors could fall into the high risk category, with all the regulatory obligations that entails (and rightly so). To understand the full legal framework surrounding AI and minors, see our article on AI and minors: what European law says.

By positioning Nolis as a conversational companion that detects and redirects, HolaNolis doesn't diagnose, assess or make decisions about anyone; that is why, in our preliminary analysis (pending external review), it is not a high-risk system under the EU AI Act. This isn't a regulatory trick; it's the natural consequence of doing things right. When your ethical design aligns with regulation, it's a sign the approach is correct.

The ethical framework: a minor's safety always comes first

At HolaNolis we follow a priority order that never changes:

1. Minor's safety: above everything else, including user experience, engagement metrics, and profitability.
2. Regulatory compliance: respecting the law is not optional.
3. Parental trust: parents need to know the tool acts in their child's best interest.
4. Teenager experience: making Nolis engaging, interesting, and useful — within the limits above.

This order demands difficult decisions. It would be more "fun" if Nolis could talk about diets or give advice on managing anxiety. It would probably increase usage time. But increasing engagement at the cost of safety is exactly what social media has been doing for a decade, and the results speak for themselves.

We don't want to be another tool that prioritizes metrics over people. Especially when those people are between 10 and 20 years old.

What parents should know about AI and health advice

If your child uses any conversational AI (not just HolaNolis), these are the key points you should understand:

• No general-purpose AI is qualified to give health advice. Not ChatGPT, not Gemini, not Claude, not Nolis. None of them.
• Disclaimers are not enough. "This does not replace medical advice" at the end of a 1,000-calorie diet plan doesn't protect your child from following it.
• Speed and availability are a risk. At 3am, AI is the only available listener. And at that hour, judgment is at its lowest.
• Talk to your child about the difference between venting to an AI (healthy) and asking it for health solutions (dangerous). For practical conversation guides, read how to talk to your child about digital safety.

When to seek professional help

Not everything a teenager tells an AI requires intervention. But there are clear signs that it's time to seek professional help:

• Persistent changes in sleep, appetite, or energy levels (more than two weeks)
• Social withdrawal with no specific cause
• Expressions of hopelessness ("nothing makes sense," "it doesn't matter what I do")
• Self-harm or mention of self-harm, in any context
• Abrupt behavioral changes (aggression, frequent crying, withdrawing from activities)
• Obsession with weight, food, or body image

If HolaNolis detects any of these signals, it will activate the alert protocol. But don't wait for an AI to tell you: if you observe these changes, act. Your pediatrician, a child and adolescent psychologist, or specialized helplines are the next step.

Frequently Asked Questions

In summary

Nolis is designed not to give medical or psychological advice because no AI should. Not because it technically can't, but because doing so would be irresponsible, dangerous, and contrary to everything HolaNolis stands for.

What Nolis does offer is a combination designed for teenagers: a space to express yourself that points you back to the trusted adults in your life, a risk detection system, alerts that reach the family immediately when there are signs of serious risk, and a constant redirection toward the professionals who can actually help.

Because the best AI for teenagers isn't the one that has all the answers. It's the one that knows when it shouldn't answer. If you'd like to see how the system works in practice, check out the quick start guide or visit the product section.