Read this first. Everything here is for understanding and organisation. It is not medical advice and does not replace a doctor, pharmacist, nurse or any other qualified professional. Never use an AI chat to diagnose yourself, to decide on a medicine or a dose, or to interpret a test result in place of a clinician. If you have chest pain, signs of a stroke, difficulty breathing, severe bleeding or any symptom that frightens you, contact your local emergency services rather than a chat window.
The most valuable thing an AI chat can do for your health is help you ask better questions, not answer them. That sounds like a disclaimer and it is actually the method: the prompts below are built to produce questions, summaries and plain language, and to refuse the thing you will be tempted to ask for. The reason that matters is specific, and it is the next section.
Why a chat is dangerous about health in one particular way
It is not that models are stupid about medicine. They are fluent about it, which is the problem. Three mechanisms cause nearly all the harm.
It agrees with you. Ask "could this be something serious?" and you get a careful, well-written case for something serious. Ask "is this probably nothing?" about the same symptom and you get a reassuring answer instead. Both replies are produced by the same machine from the same facts, steered entirely by how you phrased the question. If you are anxious, you will phrase it the anxious way, and the answer will sound like confirmation from something that reads as expert. That is the loop behind a lot of AI-fuelled health worry, and it is also how a genuinely serious symptom gets talked away.
It cannot see you. No examination, no vital signs, no bloods, no history, no idea what you look like today compared with a month ago. A clinician's judgement is built substantially on those, and none of them are in the chat.
It is confidently specific about numbers. Reference ranges, dosages, interactions and incidence rates come back in a clean, authoritative format and some of them are wrong. In any other subject that is an annoyance.
The fix is not to avoid the tool. It is to use it only for tasks where those three failure modes cannot hurt you, and to write the prompt so it cannot drift into the ones where they can.
The line, drawn precisely
| Task | Verdict | Why |
|---|---|---|
| Turning a leaflet, letter or consent form into plain language | Safe and genuinely useful | It is translation. You still take the meaning back to the person who wrote it |
| Organising what to raise at an appointment | Safe and genuinely useful | The output is your questions, not its answers |
| Writing down what you were told afterwards, so you can check it | Safe and underused | Most people forget most of a consultation. This is the half nobody does |
| Appointments, repeat prescriptions, insurance admin, records | Safe | Pure organisation with no clinical content at all |
| General wellbeing structure: sleep, movement, routine | Safe when kept general | Anything specific to your body or condition goes to a professional |
| Asking what a symptom might be | No | It agrees with the direction you asked from, and it cannot examine you |
| Interpreting a test result, scan or reference range | No | A number without your history and examination is not an interpretation |
| Anything about a medicine, a dose, a change or an interaction | No | Your pharmacist does this for free, accurately, and knows what else you take |
| Deciding whether to seek care, or whether something can wait | No | This is the decision most likely to be steered by how you asked |
| Anything urgent, or anything about a child's illness | No | Contact a professional or emergency services, immediately |
What not to paste
A separate question from safety, and one people answer too late. A consumer AI chat is a third-party service, not a clinic. Your full record, a scan, an insurance claim or a letter naming your diagnosis is sensitive data about you, and once it is pasted, it is pasted. Before you paste anything real, take the names, dates of birth, record numbers and addresses out - and consider whether you need to paste the condition itself at all, since most of these prompts work perfectly well on a description with the identifying detail removed. If the task is understanding a leaflet, the leaflet is public and there is nothing to redact.
Prompt 1 - plain language, with the diagnosis door closed
The guardrails in this one are not politeness. They stop the conversation from turning into the thing that hurts.
Prompt 2 - arrive prepared, in the ten minutes you get
A consultation is short and the thing you most wanted to say is the thing you forget. This produces a single page you can hand over or read from.
Builds a general routine around the week you actually have - sleep, movement, breaks - rather than a template you abandon by Wednesday. It holds the non-clinical boundary by design: it will not discuss symptoms, medicines or results, and says so rather than drifting.
View Tamsin - Wellness Coach →Prompt 3 - the half nobody does: after the appointment
Most of a consultation is forgotten within a day, and what survives is often the part that worried you rather than the part that mattered. Write down what you remember while it is fresh, then have it turned into something you can check.
Prompt 4 - health admin, with nothing clinical in it
This one has no safety problem at all, which is why it is the most immediately useful prompt on the page for anyone managing care for themselves or a relative.
General wellbeing, kept general
Sleep, movement, breaks and routine are the part of wellbeing where a chat genuinely helps, because the difficulty is structure rather than knowledge. Keep it general: a routine that fits your week is a planning problem, while anything that touches a condition, an injury, a medicine, pregnancy or a restrictive diet is not, and goes to a qualified professional. Nutrition in particular is covered separately, along with why the calorie numbers a model produces are guesses, in the nutrition coach guide.
If you work in healthcare, you want the other page
This guide is written for people managing their own health or a relative's. Clinical documentation is a different job with different rules, and the catalogue keeps them apart rather than blurring the line: see the medical scribe guide for encounter notes, the medical writer guide for regulatory and patient-facing documents, and the pharmacist and allied health guides for those roles. Those pages deal with the confidentiality rules, HIPAA and GDPR included, that apply the moment real patient data is involved.
The clinician-facing half: clinic letters, case summaries, discharge and referral correspondence, patient education text, all drafted for professional review before anything is sent. Not for personal health use, and the confidentiality question applies before you paste anything real.
View Healthcare Documentation Prompt Library →Where to stop, in one line
If the next thing you want to type would be answered by a person with a licence, stop and ask that person. Symptoms, results, medicines, doses, whether to seek care, anything involving a child, anything urgent: those are not slow because the system is inefficient, they are slow because judgement is involved and the judgement is not in the chat. If you are struggling with your mental health, that is also a conversation for a professional or a helpline in your country rather than a chat window, and it is worth having sooner than feels necessary.
Everything else on this page - the reading, the preparing, the remembering, the admin, the routine - is yours to automate, and doing it well is what buys you a better ten minutes when you do see someone.
For everyday life, not the clinic
The skills that fit the non-clinical side of this page, each $19 and each a one-time download:
- Tamsin - Wellness Coach - general routine and habit structure
- Petra - Fitness & Wellness Planner - movement planning around a real week
- Freya - Nutrition & Diet Coach - general eating structure, with the numbers treated as estimates
The wider sets are personal & lifestyle skills and life admin skills, and for clinical work the healthcare, science & education skills.
In summary:
Use an AI chat for understanding, preparing, remembering and admin, and for nothing that a licensed person would answer. The specific danger is that it agrees with the direction you asked from, so write the prompt to produce questions rather than answers, and take the identifying detail out before you paste. For the general wellbeing side, Tamsin - Wellness Coach AI Skill ($19); for clinicians, the Healthcare Documentation Prompt Library ($9). Both work in Claude, ChatGPT and any AI chat, with a 30-day money-back guarantee. Not medical advice.
Tamsin - Wellness Coach AI Skill
One file for the general wellbeing side: a routine built around the week you actually have, that holds the non-clinical line by design rather than drifting into symptoms, medicines or results. No subscription. Yours permanently.
KissMySkills is a marketplace of 2,000+ AI skills, 300+ prompt packs, 100+ agents & free tools for Claude, ChatGPT & any AI chat.
Related Guides
- How to Use Claude as a Nutrition Coach (and Why Its Calorie Numbers Are Guesses)
- How to Use Claude as a Medical Writer: What It Can Draft
- How to Use Claude as a Medical Scribe: The Ethan Skill Guide
- How to Use Claude as a Pharmacist's Documentation Aid: The Catherine Skill Guide
- How to Use Claude as an Allied Health Assistant: The Lucas Skill Guide
- How to Use Claude for Mental Health Documentation: The Benjamin Skill Guide
