THE 3-STAGE RESEARCH SYSTEM
Most people use AI research wrong. They ask one question and accept the answer. Here's the system that actually works:
Stage 1 — DISCOVER: Use Perplexity to get oriented on a topic Stage 2 — DEEPEN: Use Claude to analyse specific sources and documents Stage 3 — VERIFY: Cross-reference key claims with primary sources
STAGE 1: PERPLEXITY PROMPTS FOR DISCOVERY
Use these at the start of any research project:
"Give me an overview of [TOPIC] including the key debates, main perspectives, and most important recent developments. Cite your sources."
"What are the 5 most important things I need to understand about [TOPIC] if I'm coming to it fresh?"
"What do experts disagree about in [FIELD/TOPIC]? What are the main camps?"
"What are the most credible sources I should read to understand [TOPIC] properly?"
STAGE 2: CLAUDE PROMPTS FOR DEEP ANALYSIS
Once you have sources, use Claude to process them:
"I'm going to paste an article about [TOPIC]. After reading it, give me: 1) the main argument in one sentence, 2) the three strongest points, 3) any weaknesses or gaps in the argument, 4) what questions remain unanswered."
"Summarise this document in 200 words, then list the 5 most important facts or claims it makes." [paste document]
"Compare these two perspectives on [TOPIC] and tell me where they agree, where they disagree, and which arguments you find more persuasive and why." [paste both]
"What's missing from this analysis? What perspective or evidence would strengthen or challenge it?" [paste text]
STAGE 3: VERIFICATION FRAMEWORK
Before trusting any AI-generated fact, ask:
- Can I find this confirmed in a primary source?
- Is this the kind of fact that changes frequently (statistics, prices, rankings)?
- Did the AI give me a source I can check?
Red flags in AI research:
- Specific statistics without a named source
- Recent events (AI knowledge cutoffs are real)
- Legal, medical, or financial specifics
- Quotes attributed to real people
DOCUMENT ANALYSIS WORKFLOW
For analysing PDFs, reports, or long articles:
- Upload to Claude or use ChatPDF
- Start with: "Before I ask you questions, give me an executive summary of this document."
- Then ask specific questions: "What does this document say about [specific topic]?"
- Ask for gaps: "What does this document NOT cover that would be important to know?"
- Ask for plain English: "Explain the key finding on page 12 in simple terms."
BUILDING A RESEARCH DATABASE
Use Notion with Notion AI to build a searchable research library:
- Create a database with fields: Topic, Source, Key Claims, Credibility, Date
- Paste in notes from each research session
- Use Notion AI to tag and categorise entries automatically
- Ask Notion AI to surface related notes when starting new research
PROMPT: LITERATURE REVIEW GENERATOR
"I'm researching [TOPIC] for [PURPOSE]. I'll give you 5 key sources I've found. After reviewing them, write a 300-word literature review that synthesises the main themes, identifies agreements and tensions between sources, and flags what questions remain open."
Downloaded from DavideGPT.ai — Follow @DavideGPT for daily AI tips