How do you ask good questions about a field you don't know?

Good prompting advice usually says "be specific." That is hard to follow when you are new to a subject, because you don't yet know which details matter. If you have never shopped for a heat pump, you can't tell whether your climate, your ducts or the size of your house is the thing that will change the answer. A vague question gets a generic reply, and you can't spot what's missing.

The way out is to flip the order. Instead of asking for an answer first, ask the AI to act as an interviewer: what would a specialist need to know before advising you? You answer those questions, and the conversation that follows is built on your real situation. This fits a principle in A Beginner's Guide to ChatGPT: the quality of an answer depends on the context you supply. When you don't know what context to give, the tool can help you find it.

Step 1: Ask for the expert's questions

Give a role, your goal and your level, and ask for questions rather than advice.

Act as an experienced, independent home-heating advisor. I'm a homeowner thinking about replacing my oil furnace with a heat pump, and I know almost nothing about them. Before giving any advice, list the 10 questions you would ask me to understand my situation. For each question, add one sentence on why the answer matters. Use plain language.

An illustrative reply might cover your climate and coldest winter temperatures, the size and age of the house, insulation and windows, whether you have ducts, your current heating costs, what backup heat you have, and your budget. The exact list varies, but notice what has happened: you now have a map of the factors an expert considers.

Step 2: Answer them, and ask for the gaps

Copy the questions into a document and answer each one in a line or two. Where you don't know an answer, say so; that is useful too.

Here are my answers to your questions: [paste answers]. Based on these, what are the two or three most important decisions I face? Which of my "don't know" answers matter most, and how could I find them out?

This turns a fuzzy goal into specific decisions, such as "air-source or ground-source?" or "keep the furnace as backup?", and tells you what homework to do first.

Step 3: Learn the vocabulary as you go

Every field has terms that unlock better questions. Ask for them directly, in context.

List the technical terms I'm likely to see in heat pump quotes and brochures. For each, give a one-sentence definition and an example of a question I should ask an installer that uses the term.

For heat pumps, you'll meet terms like these. The definitions here are from Natural Resources Canada's consumer guide.

Step 4: Check and take it to a person

AI tools can state wrong details with confidence. Treat the conversation as preparation, not as the final word. Check numbers, rebates, rules and product claims against official sources, such as government energy agencies, and use your new questions with a qualified installer. You'll get more from that meeting because you can now follow, and test, what they tell you.

How do you find out what you should be asking?

Ask about the shape of the field, not just your problem. Useful prompts include: "What are the most common mistakes beginners make when choosing X?", "What do experienced people disagree about?", and "What question do people usually forget to ask until it's too late?" Each one surfaces questions you didn't know existed.

Ask for the skeptic's view. "What would an independent expert want to double-check in a salesperson's quote?" brings out the questions that protect you, not just the ones that inform you.

Compare with a trusted outline. Official guides often reveal the important topics through their structure alone. The Natural Resources Canada guide, for instance, gives whole sections to sizing and to supplementary heating, a strong hint that those are decisions you shouldn't skip. If a topic appears in the guide but never came up in your AI conversation, ask about it.

Finally, say what you already know and ask what you're missing. OpenAI's own prompting advice favors being specific about context and the outcome you want; when you don't yet know enough to be specific, telling the model your starting point lets it point to the next question worth asking.