Nourish, Then Flourish

Julia Brosnan Founder Tree of Health Herbalist Metabolic Balance Buteyko Breathing

Can AI Replace Your Health Practitioner?

I want to start by saying something you might not expect from a practitioner writing this post.

AI is genuinely useful.

I use it myself. It synthesises research quickly, explains mechanisms clearly, and surfaces questions worth asking. I am not here to tell you it is a fad, or that it has no place in health research. That would not be true, and it would not be helpful to you.

What I am here to tell you is that there is a gap between what AI can do and what most people think it can do. And in health, that gap carries real consequences.

What AI actually does well

When you type a symptom into ChatGPT or ask it to explain a condition, it draws on an enormous bank of published information. It can explain the mechanism behind histamine intolerance, outline the research on a particular nutrient, or list the herbs traditionally used for a given complaint, often in clear, accessible language that a textbook would not manage.

That is genuinely valuable. For understanding a diagnosis you have already received, for researching a topic before an appointment, or for making sense of a test result, AI is a useful tool.

It is also available at 11pm when you cannot sleep because you have a chronic symptom you can’t shake off, and you want to understand what might be happening. I understand the appeal completely.

Where it creates the illusion of a complete answer

Here is the problem: AI gives you information. It does not give you a clinical picture – those two things are not the same.

When you describe your symptoms to an AI, it pattern-matches against probability. It gives you the most statistically likely explanation for the words you have typed. What it cannot do is see you, ask the follow-up question your answer just prompted, notice the thing you did not think to mention, or hold your full history in view while it reasons.

I will give you a real example from my own life. Earlier this year I used AI to help me think through my own chronic sinus congestion. It gave me a plausible, well-reasoned picture connecting histamine, oestrogen, gluten and mould. That picture was useful and it was largely accurate.

But it did not know I had been living in a mouldy house. It did not know I’m in perimenopause or my cycle history, my stress load at the time, or what my diet was like and my gut had been doing for the six months prior. That was the kind of information that a practitioner knows, as it is hugely relevant for a good clinical outcome.

Most people do not know what information is relevant to supply – this is not a criticism. It is simply that knowing which details matter, and in what combination, as in collection of symptoms and whole person picture, is what clinical training is for.

The danger is not that AI gives wrong answers. It is that AI gives answers that sound complete, when they are actually built on a fraction of the information a proper assessment would use. That is the illusion of a complete answer, and it is what leads people to act on information that was never designed to account for their specific situation.

The particular risk with herbal medicine

Herbs are not supplements in the way most people think of supplements. They are pharmacologically active. They interact with medications. They have contraindications. Some are inappropriate during pregnancy. Some should not be combined with certain conditions. Some, used correctly and in the right context, are extraordinarily effective. Used incorrectly, they can cause harm or, more commonly, simply do nothing, mostly because the wrong part of the plant was used, or a different herb would have been better to use than the one that was initially chosen, or the herb was used at the wrong stage of their condition.

The research backs this up. Hepatotoxic (liver damaging) effects have been reported for more than 300 plant species. Herb-drug interactions range from negligible to life-threatening. Documented risks of herbal self-medication include incorrect self-diagnosis, severe adverse reactions, and dangerous drug interactions – particularly in people taking prescription medications, older adults managing multiple conditions, and pregnant women. A review published in Frontiers in Pharmacology found that the majority of herbal products are self-prescribed with inadequate knowledge of adverse reactions, contraindications, and interactions with existing pharmaceuticals. This is not a fringe concern. It is a documented patient safety issue.

When someone asks AI what herb to take for their anxiety, or their hormones, or their gut, and then orders that herb online and begins self-prescribing, many crucial things have been skipped: A full case history. An understanding of what else they are taking. An assessment of which tissue state or pattern they are actually presenting with, because two people with identical symptoms can have entirely different underlying drivers and require entirely different herbs to address them. An understanding of dose, form, and duration. And the clinical experience to know when a herb is helping, when to adjust, and when something else is going on entirely.

The liquid herbal extracts I use in clinic are not available over the counter. That is not an accident. They are accessible only through a qualified herbalist or naturopath because the practitioner-patient relationship and the clinical assessment that precedes prescribing are part of what makes them appropriate and effective. What is available without a practitioner is typically something that is not therapeutically active enough and is effectively a lower-dose, or it is a single-herb product, or a product a combination of herbs and/or nutrients. None of these may or may not match what AI has recommended, and none have had the individualisation that makes herbal prescribing from a qualified practitioner work so well.

AI can tell you that Withania somnifera (Ashwagandha) has adaptogenic properties and documented evidence in supporting the stress response. What it cannot tell you is whether Withania is right for you, at what dose, in what form, alongside what else, and for how long. That assessment requires a person who has been trained to make it.

What a practitioner does that AI cannot

Clinical observation:

When you sit across from a practitioner, they are observing things you have not mentioned. Skin colour, energy in the room, the way you answer certain questions, what you hesitate on. I have had clients describe their digestion as fine, and then something they made me ask one more question that changed the entire clinical picture. AI has text. A practitioner has context.

Dynamic reasoning in real time:

A practitioner follows an unexpected thread the moment it appears. If you mention something in passing that seems unrelated to your main complaint, a trained clinician will often recognise it as the most important thing you said. AI works sequentially, answering what you asked. It does not know what you did not ask.

Holistic picture, held simultaneously with clinical experience insights:

Your hormones, your gut, your sleep, your stress, your medication list, your history, your diet, your environment. A practitioner holds all of this at once and reasons across it. AI works only with whatever is in your message.

Accountability and the therapeutic relationship:

When you are working with a practitioner, they are watching. They notice if you are not improving, adjusting the plan when life happens. A practitioner asks how the elimination trial actually went, and applies the holistic view on the results of the progress so far. That relationship is part of the clinical intervention, not a nice-to-have beside it.

Appropriate investigation:

A practitioner knows which tests to order, and why. Not just the standard GP panel, but the functional testing that looks at what the standard panel was not designed to find. And they can interpret those results against your full clinical picture, not in isolation.

Legal scope and TGA compliance:

AI will occasionally make a specific health claim or herbal recommendation that a practitioner cannot ethically make without a proper assessment, because a practitioner is accountable for what they recommend in a way that an AI is not. The absence of that accountability is not a feature. It is a risk the person consulting AI absorbs without knowing it.

Why the health fund reinstatement matters right now

For six years, from April 2019 to July 2025, Western herbalists and naturopathic herbalists were excluded from private health insurance rebates in Australia. That exclusion was based on a review of the evidence that many practitioners, myself included, considered inadequate and incomplete. The practical result was that qualified, experienced practitioners became harder for people to access and afford, and many people who would otherwise have sought a clinical assessment turned elsewhere instead.

From 1 July 2026, private health funds can once again provide rebates for consultations with registered Western herbalists and naturopathic herbalists. This is a meaningful shift. For many people, cost has been a genuine barrier to accessing qualified care. That barrier is now significantly lower than it has been for six years.

If you are privately insured and you have been putting off seeing a practitioner, it is worth checking your extras cover now. The level of rebate varies by fund and by tier, and the claims infrastructure is still rolling out across different funds, but the legislative change is in place and the direction is clear.

How to actually use AI well for your health

I am not suggesting you stop using AI. I am suggesting you use it for what it is actually good at.

Use it to understand a diagnosis you have received, to make sense of terminology in a test result, to research a condition before your appointment so you can ask better questions. Use it to understand the mechanism behind a treatment your practitioner has recommended. Use it to learn.

Do not use it to replace the assessment that should come before any clinical decision. Do not use it to self-prescribe herbs, supplements, or dietary protocols for a complex or ongoing health issue. And do not mistake a confident-sounding answer for a complete one.

The most important thing AI cannot do is know what it does not know about you. A practitioner who has taken a full case history, reviewed your history, and sat with you for an hour knows what is missing. AI works with what you give it, and what most people give it is a fraction of what is clinically relevant.

A note on this blog itself

I want to be transparent about something.

Some of the thinking in this post was developed in conversation with an AI. I find it useful for organising ideas, checking my reasoning, and making sure I have not missed something. That is a legitimate use.

The clinical knowledge, the experience of sitting with clients, the understanding of what gets missed in a text-based exchange, and the judgement about what matters in a specific person’s case, none of that came from AI. It came from years of clinical training and practice.

That is the distinction that matters. AI as a tool, in the hands of someone who knows what they are doing with it, can be genuinely useful. AI as a replacement for the person who knows what they are doing with it is where the risk lives.

Where to start

If you have been piecing together health answers from online research and AI conversations, and something still is not resolving, that is useful information. It usually means the picture is more complex than a pattern-match can capture, and that a proper clinical assessment would find something the research alone has not.

A Clarity Call is a no-pressure starting point to talk through what is going on and whether working together makes sense.

AI gives you information. A practitioner gives you a clinical picture. They are not the same thing.

References & Resources

If you’ve been trying to improve your health but feel like your body isn’t responding the way it used to, there is usually a reason.

This is exactly what I explore with clients inside my programs.