Skip to main content

Deep Research

"Use AI to investigate, not just answer."

Beyond Quick Answers​

Most AI use follows this pattern:

  1. Have a question
  2. Ask AI
  3. Get an answer
  4. Move on

But AI's real power for research is different. It can investigate, exploring a topic from multiple angles, synthesizing information, and helping you understand deeply.

Quick Answer vs Deep Research​

Quick AnswerDeep Research
"What is X?""Help me understand X deeply"
Single responseMulti-turn exploration
Takes the answerQuestions the answer
Surface knowledgeConnected understanding
Closes the loopOpens new questions

When to Go Deep​

The Right Signals​

  • You'll make decisions based on this knowledge
  • Getting it wrong has significant consequences
  • You need to explain this to others
  • The topic connects to many other things
  • You've gotten conflicting information elsewhere

The Wrong Time​

  • You just need a quick fact
  • You're in execution mode, not learning mode
  • The stakes are low
  • You'll verify through other means anyway

Deep Research Techniques​

1. Start with Orientation​

Don't jump to your specific question. Understand the landscape first.

I need to understand [topic area] for [purpose].

Before I ask my specific question, give me:
1. Key concepts I should know
2. Common misconceptions
3. How experts think about this
4. Questions I should be asking

2. Explore Multiple Angles​

Ask AI to present different perspectives.

Explain [topic] from these perspectives:
1. A practitioner who uses this daily
2. Someone skeptical of this approach
3. A teacher explaining to a beginner
4. An expert debating edge cases

3. Question the Answers​

Don't accept responses at face value.

You said [claim].
- What's the evidence for this?
- What are the counterarguments?
- When is this NOT true?
- What am I likely misunderstanding?

4. Connect to What You Know​

Build bridges to existing knowledge.

I understand [thing I know well].
How does [new topic] relate to this?
What's similar? What's importantly different?

5. Test Understanding​

Verify you actually understand, not just read.

Let me explain [topic] back to you:
[your explanation]

What did I get right? What's wrong or missing?
What's the most important thing I'm not seeing?

Research Session Structure​

Phase 1: Explore (15-30 min)​

  • Broad questions about the landscape
  • Multiple perspectives
  • Map out key concepts
  • Identify what you don't know you don't know

Phase 2: Focus (15-30 min)​

  • Dive into specific areas
  • Ask "why" and "how" questions
  • Challenge assumptions
  • Look for edge cases

Phase 3: Synthesize (10-15 min)​

  • Summarize your understanding
  • Get feedback on that summary
  • Identify remaining gaps
  • Plan next steps

Phase 4: Document​

  • Write down what you learned
  • Note remaining questions
  • Record key decisions and reasoning

Research Prompts​

For Technical Topics​

I'm researching [technology/pattern/approach].

My context: [your situation]

Help me understand:
1. Core concepts and how they relate
2. When this is the right choice
3. When this is the WRONG choice
4. Common mistakes people make
5. How this evolves over time

Let's explore this step by step.

For Decision Making​

I need to decide between [option A] and [option B].

Context: [your situation]

Walk me through:
1. Key factors to consider
2. How each option scores on these factors
3. Hidden trade-offs I might miss
4. What questions I should be asking
5. How to think about this decision

Don't tell me what to decide. Help me understand.

For Problem Investigation​

I'm trying to understand why [problem/behavior exists].

What I observe: [symptoms]
What I've tried: [attempts]
What I suspect: [hypotheses]

Help me investigate:
1. Is my framing of the problem correct?
2. What could cause these symptoms?
3. How would I test each hypothesis?
4. What might I be missing?

Signs Research Is Working​

Good Signs:

  • You're discovering questions you didn't know to ask
  • Your mental model is evolving
  • You can explain the topic better than before
  • You understand trade-offs, not just features
  • You know what you still don't know

Warning Signs:

  • You're collecting information without integrating it
  • Every answer feels final
  • You're not asking follow-up questions
  • You couldn't explain this to someone else
  • No "aha" moments

Common Mistakes​

Accepting First Answers​

The first response is a starting point, not the end. Push deeper.

Not Challenging Claims​

AI can be confidently wrong. Ask for evidence, counterarguments, edge cases.

Research Without Purpose​

Exploration is good, but know what decision or action your research supports.

Not Documenting​

Research you don't document fades. Write down key insights, even briefly.

Going Alone​

AI research is powerful, but also check authoritative sources, talk to experts, and verify independently.

AI Research vs Traditional Research​

AI research is excellent for:

  • Rapid orientation to new topics
  • Synthesizing across many sources
  • Explaining concepts at your level
  • Exploring connections and implications

Traditional research is still needed for:

  • Authoritative/citable sources
  • Latest information (beyond AI knowledge cutoff)
  • Verification of AI claims
  • Primary sources and original data

Best approach: Use AI for exploration and synthesis, traditional sources for verification and depth.