Artificial intelligence (AI) is everywhere right now. Patients and parents are using it to interpret lab results, make sense of symptoms, suggest dietary interventions, choose supplements, and even decide whether a medication is “safe.”

On the surface, this makes sense. People are seeking clarity, empowerment, and a sense of control in a healthcare system that can feel overwhelming.

But across each of our 25+ years in clinical practice, we’re noticing a concerning shift.

People are not necessarily getting sicker.
But they are getting harder to treat.

This shift didn’t happen overnight. It began with search engines like Google, intensified with social media, and has accelerated rapidly with AI. Each step has added more information, more opinions, and less context.

When individuals are exposed to constant, often conflicting streams of health information, it becomes difficult to settle into a coherent path forward. Instead of feeling supported, many feel anxious, paralyzed, or unsure of whom—or what—to trust.

In this post, we’ll explore 10 dangers we’re seeing in clinical practice with the growing reliance on AI for health information.

We are not dismissing AI’s usefulness (we use it ourselves in various ways) but rather want to highlight where it falls short and begins to contribute to confusion rather than clarity, sometimes delaying healing in the process.

1. AI Substitutes Confidence for Accuracy

AI is very good at sounding certain—even when it’s wrong!

It presents information cleanly, fluently, and authoritatively. But behind the scenes, it’s generating responses based on probability, not professional experience. It doesn’t communicate uncertainty well. It rarely says, “This depends,” in a meaningful, nuanced way.

Experienced clinical judgment lives in the gray. AI speaks in black and white.

In practice:
Patients may come in convinced they have SIBO, PANDAS, POTS, mold illness, or a specific food intolerance (all conditions that share a wide range of overlapping symptoms) because AI gave them a clean, confident answer.

The danger isn’t that AI might be wrong; it’s that it can be wrong with conviction.

2. Pattern Matching Is Not the Same as Diagnosis

AI excels at pattern recognition, while accurate diagnosis requires pattern interpretation. AI works by recognizing patterns in text, not by understanding physiology. It matches symptom clusters to common explanations. It doesn’t evaluate timing, trajectory, severity, or context. It can’t examine a patient or sense what doesn’t fit.

Two people can share the same symptoms and have entirely different underlying causes and therefore very different paths to healing.

In practice:
Two kids with anxiety and stomach pain may look identical to AI, but one has zinc deficiency and the other is dealing with the emotional stress of his parent’s recent separation.

AI flattens nuance. Good clinicians understand how to engage with it.

3. AI Doesn’t See the Body as a System

Most AI-generated health advice is reductionist by nature. It’s programmed to believe that one symptom has one cause and one fix. Add another symptom, get another fix.

The human body doesn’t work that way.

One question we ask constantly in our practices is: “What is this individual’s body trying to communicate?”

AI’s focus on fixing or optimizing often misses the reality that many chronic symptoms are signals, not errors.

In practice:
We see patients stacking supplements, elimination diets, and protocols that technically make sense in isolation, but together overwhelm digestion, nervous system regulation, or a healthy relationship with food.

Years of sitting across from people provides us with immediate feedback—facial expressions, body language, tone of voice, and direct dialogue that tell us when a person is onboard or getting overwhelmed.

AI is unable to sense when a system is becoming fragile or to protect the body as a whole. Treatment suggestions can become reactive instead of restorative.

4. AI Has Changed the Clinical Dynamic

One of the most interesting observations we’ve made is how AI has changed the clinical dynamic. Trust has subtly shifted from relationship to algorithm and clinical expertise has been demoted to another “opinion.”

Instead of visits grounded in curiosity and collaboration, conversations can drift toward debate and defensiveness. This erodes the therapeutic partnership that actually allows healing to occur.

The truth is, there is often more than one valid path to healing.

But healing requires coherence. When people are flooded with conflicting information, opposing opinions, and endless “what ifs,” they get stuck.

In practice:
We may hear: “My research found that lithium is dangerous,” without distinguishing between prescription doses and low-dose supplementation. Or “AI said if I can’t tolerate almonds, then I should avoid all oxalates.”

Without a doubt, in our experience, healing works best when there is trust, not constant second-guessing. People may have too much information to move forward confidently and trust too little to be cared for. When this occurs, even good treatments can fail to work as well as they should.

5. AI Affirms What It Thinks You Want to Hear

AI responses are shaped not only by what is asked, but by who is asking—or more accurately, by patterns in prior questions, interests, and topics of inquiry. Over time, AI learns what themes you tend to explore and begins to respond through that lens.

This means AI doesn’t just provide information; it subtly reflects back familiar frameworks, reinforcing existing ways of thinking rather than challenging them. (The same is true for Google, social media…)

In practice:
We’ve noted that if Debbie asks AI a question about anxiety, for example, the responses may point toward nervous system regulation, nutrient status, or thyroid function—areas she often explores clinically or writes about. When Dr. Roy asks a question about anxiety, AI readily suggests immune-mediated conditions like PANDAS or inflammatory drivers, reflecting the patterns and conditions he routinely investigates.

The issue isn’t that these considerations are wrong. The concern is that AI may narrow the field of possibilities too quickly, reinforcing familiar narratives rather than helping someone step back and ask, “What else could be going on?”

6. AI Can Find Something Negative About Any Treatment

No matter how safe, effective, or appropriate a treatment may be, AI can generate reasons to question it. Medications. Supplements. Protocols. Diets.

This sets the stage for hesitation and mistrust, where nothing feels safe enough to try for long enough to actually work.

In practice:
We may suggest a supplement that we have used for decades upon decades with hundreds of people, but if AI lists a minor but theoretical side effect (such as the amino acid L-glutamine can be excitatory, zinc causes nausea, or magnesium lowers blood pressure) someone may be hesitant to even take it, regardless of the form or dose, even when the risk-benefit balance clearly favors a thoughtful trial.

7. AI Can Only Answer What Is Asked

AI responses are highly dependent on how a question is framed. Patients are not trained to ask medical questions the way clinicians think through them, nor should that be expected.

A question asked without thorough and appropriate clinical context can easily produce misleading or incomplete answers. AI doesn’t know what you meant to ask, it only responds to the words that were typed.

In practice:
Two clients, who didn’t realize the subtle differences among elemental shake formulations, asked AI why they were reacting to their formula. One was told she was having a fungal reaction due to high sugar content, while the other was told she was experiencing a histamine reaction to whey protein. In the first case, the client was using an elemental shake with a glucose polymer (no sugar), and in the second, a formula with only amino acids (no whey).

8. AI Amplifies Health Anxiety

For certain people, AI becomes a 24/7 reassurance-seeking loop that never actually reassures. The more one searches, the more possibilities appear. Each new answer introduces additional “what ifs,” expanding concern rather than narrowing it.

AI can unintentionally train the brain to stay in threat mode. Instead of helping someone tolerate uncertainty, it can reinforce the idea that more searching will eventually bring relief, when in reality, it often does the opposite.

In practice:
This is particularly risky for anxious parents, teens, and individuals with perfectionistic or OCD traits, as well as anyone with prior medical trauma. Instead of calming the system, more information can keep the brain in threat mode, constantly scanning for what might be wrong.

9. AI Should Not Be Trusted with Your Mental Health

Mental health is one of the areas where relying on AI is limited at best and potentially dangerous at worst.

AI tends to label and explain rather than explore. It often affirms a person’s existing interpretation of their thoughts and feelings, reinforcing a narrative that may feel validating but isn’t always complete or helpful. Without the ability to assess context or emotional nuance, reassurance can become misleading.

In practice:
Some teens struggling with OCD traits repeatedly use AI to analyze their thoughts for reassurance. While each response may temporarily be reducing immediate stress, it also unfortunately reinforces the habit of checking and rechecking. Over time, this strengthens rumination, making symptoms more persistent and harder to treat.

10. AI Cannot Replace Relationship

At its core, a huge component of healing is relational. It unfolds through trust, continuity, and shared decision-making over time. AI can generate information, but it cannot build a therapeutic relationship, hold uncertainty, or walk alongside someone as they respond.

In practice:
One of the notable patterns we see is that teens and adults, who have access to online health information and AI-driven answers, are often harder to treat than both our young pediatric patients and our clients in their 60s, 70s and 80s.

Those who typically rely less on the internet for interpretation and reassurance are often more willing to trust a plan and stay with it. Similarly, they tend to allow someone else to understand and hold the complexity for them, often allowing their nervous system to settle and healing to take place.

This doesn’t mean information isn’t valuable. It means information works best when it is held within a trusted and caring relationship.

If you’ve read this far, we’d love to know if this resonates with you.

• Where have you found AI or online health information helpful?
• Where has it missed the mark?
• Does one of these 10 speak most strongly to you?

We’d love to hear about your experience in the comments below.