Your patients are already using AI. The question isn't whether to engage — it's how to do so safely.
The Reality
40% of UK adults have used generative AI.
Many are asking ChatGPT about symptoms, medications, and test results — often before (or instead of) consulting their GP. This framework provides practical guidance for UK general practice — not to promote AI use, but to reduce harm when patients inevitably use these tools.
The Core Problem
Patients using AI for health information face real risks:
Hallucinations
AI confidently states incorrect medical information
Delayed Care
Reassurance from AI delays seeking professional help
Wrong Context
US-centric advice (dosages, drug names, reference ranges)
Missing Red Flags
AI may not recognise urgent symptoms requiring immediate care
Key insight: Telling patients "don't use AI" is ineffective. A harm reduction approach is more realistic.
The Harm Reduction Approach
Rather than prohibition, we can guide safer use:
Lower Risk Uses
- • Understanding medical terminology
- • Preparing questions for GP appointments
- • Learning about conditions after diagnosis
- • General health education
Higher Risk Uses
- • Self-diagnosing symptoms
- • Medication dosage decisions
- • Interpreting test results alone
- • Emergency symptom assessment
Risk Ladder: Patient AI Use
From dangerous to safer — and what to do instead
"ChatGPT says I should stop my blood pressure tablets"
Changing or stopping prescribed medication based on AI advice
Potential harm
Severe
Instead: Never change medications without speaking to your GP or pharmacist first.
"I've had chest pain but Claude says it's probably just anxiety"
Using AI to decide whether symptoms need urgent attention
Potential harm
High
Instead: For chest pain, breathing difficulty, or sudden symptoms — call 999 or 111. AI cannot examine you.
"I asked Gemini what my blood test results mean"
Interpreting test results without professional context
Potential harm
Moderate
Instead: Use AI to understand what tests measure, but discuss your specific results with your GP who knows your history.
"I described my symptoms and ChatGPT thinks it might be X"
Using AI to generate possible diagnoses
Potential harm
Low-Moderate
Instead: Write down your symptoms to share with your GP, but don't anchor on AI's guess — it could be completely wrong.
"My GP diagnosed me with X — I asked Claude to explain it simply"
Learning about a condition you've already been diagnosed with
Potential harm
Low
Tip: Cross-check with NHS.uk. Ask AI to use UK terminology and guidelines.
"I used AI to help me write down questions for my GP appointment"
Preparing for consultations, understanding medical terms
Potential harm
Minimal
This works well: AI can help you articulate concerns and make the most of limited appointment time.
The principle is simple:
The higher the stakes, the more you need a human professional. AI can help you learn and prepare — but it cannot examine you, know your history, or take responsibility for your care.
What Practices Can Do
1. Acknowledge the Reality
Include AI in consultations: "Have you looked this up online or asked an AI chatbot?" This opens dialogue without judgment and lets you correct any misinformation.
2. Provide Patient Guidance
Create or share resources that help patients use AI more safely — what to check, what to avoid, when to always seek professional help.
3. Develop Practice Protocols
Consider how your practice handles AI-related queries. Document your approach for consistency across the team.
4. Staff Training
Ensure clinical and reception staff understand common AI tools and their limitations, so they can respond appropriately to patient questions.
Key Statistics
40%
UK adults have used generative AI
70%
Find it useful for health info
25%
Have used AI for symptom checking
60%
Want NHS guidance on AI use
The Framework Includes
- Risk assessment matrices — Categorising AI use by risk level
- Patient handouts — Ready-to-use guidance materials
- Staff protocols — How to handle AI-related consultations
- Governance templates — Documentation for practice policies
- Red flag recognition — When AI use becomes dangerous
Download the Complete Framework
The full 30-page document includes detailed protocols, patient resources, and implementation guidance.
Download PDFA note on how this was made
This framework was developed collaboratively using multiple AI tools. Claude (Anthropic) drafted the initial content, ChatGPT (OpenAI) provided critical analysis and refinements, and the whole was reviewed by Dr Pasupathi to ensure clinical accuracy and practical applicability for UK general practice. We believe transparency about AI involvement reflects the same principles we advocate in this document.
The goal is not to make patients dependent on AI, nor to make them afraid of it. It's to help them — and us — navigate this new reality safely.
Dr Krishnan Pasupathi
MBBS MBA MRCGP
NHS GP Partner with 29 years in medicine. GP Trainer since 2016. Building free patient education tools and resources for healthcare professionals.