When Apple Watch began detecting atrial fibrillation in 2018, it saved documented lives. When people started using ChatGPT to diagnose chest pain in 2024, it missed heart attacks. AI health tools occupy a wide spectrum of usefulness and danger — and the difference between a tool that helps you and one that harms you often comes down to how you use it, not which app you choose.
The Two Fundamentally Different Types of AI Health Tools
Understanding this distinction prevents most misuse:
Type 1 — Monitoring tools: These collect and analyse objective physiological data over time. Wearables (Apple Watch, Fitbit, Whoop, Garmin) track heart rate, sleep stages, SpO2, and activity. AI analyses patterns. These tools are genuinely valuable for trend monitoring and anomaly detection.
Type 2 — Conversational AI for symptoms: These take your verbal description of symptoms and return information. ChatGPT, Claude, and symptom checker apps (Ada, K Health, Babylon) fall here. These tools provide information — they do not provide diagnosis.
The mistake most people make is treating Type 2 tools as if they were Type 1 — as if the information they return has the same reliability as measured data.
What Monitoring Tools Do Well
- Sleep tracking: Consistent data over weeks reveals patterns — chronic poor sleep, sleep apnoea indicators, the impact of alcohol on sleep quality. This information is genuinely useful to share with your doctor.
- Heart rate variability (HRV): HRV trends over time are a well-validated indicator of recovery and stress. Whoop and Garmin devices track this effectively.
- ECG screening: Apple Watch’s ECG feature has received FDA clearance for atrial fibrillation detection. This is not diagnosis — but it is a legitimate screening tool that has prompted life-saving medical consultations.
- Glucose monitoring: Continuous glucose monitors (CGMs) like Dexcom and Abbott Libre, now increasingly paired with AI coaching, provide real-time data valuable for both diabetics and metabolically curious individuals.
The Right Way to Use Symptom AI
Conversational AI is useful for health education, not health diagnosis. The right way to use it:
Good use: "Can you explain what elevated AST levels in a liver panel typically indicate?
What questions should I ask my doctor about these results?"
Risky use: "I have had chest tightness for two days and I am slightly short of breath.
Do I need to go to the hospital?"
The first prompt seeks education to prepare for a medical conversation. The second asks AI to make a triage decision that requires physical examination, vital signs, and clinical judgment. GPT-4 has demonstrated impressive medical knowledge on standardised tests — but standardised tests do not involve patients with atypical presentations, multiple conditions, or the full complexity of a real consultation.
Red Lines: When to Call a Doctor, Not Ask AI
- Chest pain, pressure, or tightness — any severity
- Sudden severe headache unlike any you have had before
- Shortness of breath at rest
- Neurological symptoms: numbness, vision changes, speech difficulty
- Any symptom that is getting worse over 24-48 hours
- Symptoms in a child under 2 years old
For these situations, AI is not a triage tool. It is a distraction. Call your doctor, urgent care, or emergency services.
How to Use AI to Prepare for a Medical Appointment
This is where conversational AI genuinely adds value. Before your next appointment, use this prompt:
I have an appointment with my doctor about [condition/concern].
My symptoms are: [describe them]
I am currently taking: [list medications]
Help me:
1. Explain my symptoms clearly and precisely
2. List the 5 most important questions I should ask
3. Tell me what tests or examinations I should ask about
4. List any information or records I should bring to the appointment
Patients who arrive at appointments prepared ask better questions and leave with better outcomes. AI is excellent at helping you prepare for a human medical conversation — not replace it.
Apps Worth Knowing in 2026
- K Health — AI symptom assessment reviewed by physicians; lower risk than solo AI use because a doctor reviews the output
- Whoop 5.0 — best recovery and HRV tracking for performance-focused users
- Levels — CGM-based metabolic health coaching with AI insights
- Apple Health with AI summaries — iOS 18 consolidates health data from all sources into readable trend summaries
Key Takeaway: Use monitoring AI for tracking trends over time and preparing better conversations with your doctor. Never use conversational AI to triage acute symptoms. The distinction between these two use cases is the difference between a genuinely useful health tool and a dangerous one.
Be the first to respond