Use AI as a Research Assistant

This playbook teaches you how to use AI to research any topic and produce a summary you can actually trust. You'll learn the critical skill of evaluating AI output rather than accepting it at face value.

info

This content was developed with AI assistance and is regularly reviewed for accuracy.

What You'll Accomplish

• Research a topic using AI to gather and organize information
• Evaluate accuracy by spotting claims that need verification
• Fact-check key details using a simple validation process
• Produce a reliable summary with confidence levels noted

Prerequisites

• An account on any AI chatbot (Claude, ChatGPT, or Gemini)
• A topic you genuinely want to learn about
• A web browser for fact-checking (Google, Wikipedia, or relevant sources)
• 20-30 minutes

The Playbook

Step 1: Define Your Research Question

Goal: Turn a vague interest into a specific, answerable question.

Action: Start by telling the AI what you want to know and why. Vague topics produce vague answers — a focused question gets you usable information.

Example:

I'm considering switching to an electric vehicle and want to understand the
real-world costs. What are the total ownership costs of an electric vehicle
versus a gas car over 5 years, including purchase price, fuel/charging,
maintenance, insurance, and tax incentives? Focus on mid-range sedans
available in the US.

Checkpoint: Your prompt asks a specific question with clear scope and parameters.

Step 2: Ask for Sources and Reasoning

Goal: Get the AI to show its work so you can verify claims.

Action: After the initial response, ask the AI to separate facts from estimates and identify where its information comes from.

Example:

For each cost figure you mentioned, indicate whether it's based on published
data, industry averages, or an estimate. Flag any numbers that may be outdated
(more than 12 months old).

This forces the AI to be transparent about confidence levels. Pay attention to phrases like "approximately," "estimates suggest," or "as of [year]" — these signal areas to verify.

Checkpoint: You can see which claims are well-supported and which are the AI's best guesses.

Step 3: Verify the Important Claims

Goal: Confirm the facts that would actually influence your decision.

Action: Pick the 3-5 most important claims from the AI's response and verify them independently. You don't need to check everything — focus on numbers, dates, and specific facts that matter for your use case.

Verification methods:

• Search the specific claim — paste a key phrase into Google and see if reliable sources confirm it
• Check official sources — government websites, manufacturer specs, industry reports
• Look for recency — make sure the data is current, especially for prices and regulations

Example: If the AI says "the average EV saves $1,000 per year in fuel costs," search for "average EV fuel savings per year" and compare what you find against the AI's claim.

Checkpoint: You've verified the key claims and noted any corrections or updates.

Step 4: Ask for the Other Side

Goal: Get a balanced view by explicitly requesting counterarguments.

Action: AI tends to give you the answer it thinks you want. Push back to get a more complete picture.

Example:

What are the strongest arguments against buying an electric vehicle right now?
Include practical downsides that EV advocates tend to downplay.

Checkpoint: You have both the supporting and opposing perspectives on your topic.

Step 5: Compile Your Summary

Goal: Produce a final document you can reference and share.

Action: Ask the AI to consolidate everything into a structured summary that includes your corrections.

Example:

Based on our full conversation, create a summary document with these sections:
1. Key findings (bullet points)
2. Cost comparison table (EV vs gas over 5 years)
3. Main advantages and disadvantages
4. What I verified and what still needs checking
Keep it under 500 words.

Copy this summary into your own document. Add any corrections you found during fact-checking.

Checkpoint: You have a concise, verified summary saved outside the AI platform.

Common Pitfalls

• Treating AI output as sourced research — AI generates plausible text, not cited research. Always verify specific facts before sharing them as true.
• Accepting the first framing — AI may present a topic from one angle. Asking "what am I missing?" or "what's the other side?" gives you a more complete picture.
• Skipping verification on numbers — Statistics, dates, and specific figures are where AI is most likely to be inaccurate. These are also the claims that matter most.
• Using AI for breaking news — AI training data has a cutoff date. For recent events, use news sources directly.

Key Takeaways

• Specific questions with clear scope produce the most useful research output
• Ask AI to separate facts from estimates — then verify the facts that matter
• You don't need to check everything, just the 3-5 claims that would change your conclusion
• Always ask for counterarguments to get a balanced view
• Save a final summary with your corrections included

Next Steps

Try Draft Better Emails and Documents with AI to apply what you've learned to practical writing tasks, or explore the AI 101 Foundations track for a deeper understanding of how AI works.