Practice12 min read

Asking Better Research Questions

A research question isn't just a question

It's a brief. A research question worth asking AI has five dimensions:

Topic — what you're researching

Scope — what's in, what's out

Criteria — what makes something relevant or recommended

Format — how you want findings presented

Confidence flagging — explicit permission for AI to say "I'm not sure"

Leave any of these out and AI guesses. AI guesses are confident — and often wrong.

The research prompt template

I need to research [topic] for [purpose]. My scope: [what to include, what to exclude]. The criteria that matter most: [3-5 specific things]. Present findings as [format]. For any claim you're not highly confident in, flag it with "⚠️" so I know to verify.

The last sentence is the most important. Without it, AI presents weak guesses with the same confidence as strong facts.

The confidence flag works. Modern AI tools will mark uncertain claims when you ask them to. It's not perfect — they still miss things — but you'll catch the majority of hallucinations this way, which is enough to prevent most embarrassing mistakes.

Practice Exercise

Your manager asks you to research whether you should attend an upcoming industry conference. The conference is "GenAI for Business 2026" in Chicago, October 14-16. It costs $1,800 to attend. Your team has been told to focus on practical AI deployment, not theoretical content. You need to come back with a recommendation by Friday. Write the prompt that gets AI to do real research on this, not just describe the conference website.

No pressure — just give it your best shot! Write a prompt for the scenario above and our AI will give you friendly, specific feedback on how to improve.

0 / 5,000

(try writing your own prompt first!)