Prompt Engineering for Emergency Preparedness: Asking AI the Right Way

You can ask an AI tool for emergency preparedness help and get back a generic, unhelpful response. Or you can ask the same tool a slightly different way and get a personalized plan you can actually use. The difference is how you frame your question—what experts call "prompt engineering."

Think of it like asking a friend for advice. If you say "What should I do in an emergency?" they might give you vague generalizations. But if you say "My family is me, my partner, and two kids ages 4 and 8. We live in an apartment in a flood zone. My partner works across town. What should we actually plan for?" they can give you relevant guidance. AI works the same way.

The Structure of a Good Safety Prompt

An effective prompt gives the AI context about your specific situation. The best safety prompts include:

• Your situation: Where you live, what natural hazards are in your area, your housing type (apartment, house, rural), your family composition
• Your constraints: Budget limitations, mobility issues, health conditions, dependent ages, transportation access
• Your desired outcome: Not "help me prepare" (too vague) but "create a specific checklist," "design an evacuation plan," or "identify gaps in my current plan"

Bad prompt: "How do I prepare for disasters?" Good prompt: "I live in [city] in a second-floor apartment with my elderly mother and two young kids. I work 30 minutes away. Create a specific evacuation plan including how my mother moves with her walker, where we shelter, and how we reunite."

Why This Matters

Vague prompts return vague answers because the AI doesn't have enough information to personalize. It defaults to generic advice that fits nobody perfectly. When you provide specific context, the AI can tailor its response to your actual life.

You can also layer prompts—ask a follow-up question that builds on the previous answer. "Now design a communication plan for that family" or "What gaps remain in this plan?" This iterative approach lets you build a comprehensive safety strategy conversation by conversation.

Practical Techniques

One powerful technique is asking the AI to think through your constraints: "Given that we can't evacuate by car due to limited mobility, and we can't stay in shelters that are loud and overstimulating, what shelter options exist for us?" This forces the AI to work within your reality instead of suggesting generic solutions.

Another is asking for decision trees: "Create a decision tree for my family: If flooding starts, do we evacuate immediately or shelter in place? What triggers each decision? What happens next?" This gives you a plan that works even when you're panicked or unclear.

You can also ask the AI to challenge your assumptions: "I think my family's main emergency risk is earthquakes, but I live in [area]. What am I probably overlooking?" This helps you build plans for real risks, not the ones you assume.

Refinement Through Feedback

If the first answer misses something, tell the AI: "This is helpful but doesn't account for my partner's night shift or that we have three dogs." Then ask again. The AI learns your constraints as you go, and answers improve with each clarification.

Try this: Pick one specific emergency (power outage, evacuation, shelter-in-place). Write out your family situation in detail (ages, abilities, health needs, location, work arrangements). Now ask Claude or ChatGPT: "Here's my family situation. Walk me through exactly what would happen in [emergency]. What would we do in hour one, hour four, and day one?" The detail in your question will directly improve the detail in the answer.