For EMT / Paramedics ·
What you'll accomplish
By the end of this guide, you'll have a repeatable method for running interactive clinical scenarios with ChatGPT or Claude — where the AI plays both the patient and dispatch, responds to your treatment decisions with updated vitals and patient status, and gives you feedback on your clinical reasoning. Free clinical simulation, available at 2am during a slow 24-hour shift.
What you'll need
Go to chatgpt.com or claude.ai and log in or create a free account. For clinical simulations, both work well. Claude is slightly more reliable for maintaining consistent patient status across a long scenario; ChatGPT is more creative with scenario variety.
Copy and paste this prompt to start a simulation:
You are a clinical simulation system for an EMT/paramedic. Here's how we'll work:
1. You will play both the DISPATCHER and the PATIENT
2. Start by giving me a dispatch notification (nature of call, address type, patient age/sex if known)
3. When I arrive on scene, describe what I see and present the patient with initial vitals and chief complaint
4. As I perform assessments and interventions, update the patient status and give me new vitals
5. If my treatments are correct, the patient improves; if incorrect or delayed, the patient deteriorates appropriately
6. If I ask the patient questions, answer as the patient would
7. When the scenario ends (transport, ROSC, or death), give me a debrief: what I did well, what I should have done differently, and what the diagnosis/outcome was
Start the scenario now. Generate a realistic ALS-level scenario appropriate for a paramedic.
Respond to the scenario as you would on a real call. Be specific about your actions: "I apply the cardiac monitor and acquire a 12-lead" is better than "I check the heart." The more specific your actions, the more useful the AI's feedback.
What you should see: A realistic patient presentation that evolves based on your interventions.
Example response to a chest pain scenario: "I perform a scene size-up — scene is safe, no hazards. I approach the patient. I perform a primary survey: airway patent, breathing present but shallow, circulation — radial pulse present, skin diaphoretic and pale. I apply supplemental oxygen at 4 liters via nasal cannula. I place the patient on the cardiac monitor. What does the monitor show?"
Work the scenario to completion. The AI will update patient status after each intervention. If you make a clinical error (wrong medication, missed critical finding), the patient will deteriorate — creating real learning pressure.
When the scenario ends, if the AI doesn't automatically debrief, ask: "Please give me a full debrief of this scenario — what was the diagnosis, what did I do well, what should I have done differently, and what's the learning point?"
What you should see: A structured analysis of your clinical performance, the correct diagnosis, and the key teaching point for this scenario type.
Customize these to target your specific weak areas:
Cardiac arrest simulation: "Simulate a cardiac arrest scenario. Begin with a witnessed collapse. The initial rhythm will be shockable. Include at least one ROSC and post-ROSC management phase."
Stroke simulation: "Simulate a stroke call. Present with a patient who has subtle findings that could also be hypoglycemia or a complex migraine. Include a blood glucose result. Begin with dispatch."
Multi-system trauma: "Simulate a high-mechanism trauma scenario — motor vehicle collision. Include at least two distinct injury systems. Begin with dispatch to the scene."
Obstetrics emergency: "Simulate an OB emergency — a patient in active labor with a complication. I'm the paramedic on scene. Begin with dispatch."
Toxicology: "Simulate an overdose call with an unclear ingestion history. Include vital sign abnormalities consistent with a specific drug class (you choose). I must identify the toxidrome. Begin with dispatch."