Practical 10 min read

Using AI for Research & Analysis

AI can be a powerful research assistant — summarising documents, finding patterns, and helping you understand complex topics. But you need to know its limitations. Here's how to research effectively with AI.

The Golden Rule: Trust but Verify

AI models can and do make things up. They can cite books that don't exist, invent statistics, and state falsehoods with complete confidence. Never treat AI research output as fact without independent verification — especially for dates, statistics, quotes, and scientific claims.

Best Tools for Research

Perplexity AI

Purpose-built for research. Always cites sources. Best for fact-checking and quick research.

Gemini Deep Research

Browses hundreds of websites and produces multi-page reports with citations. Best for comprehensive research.

Claude (Long Context)

Excellent at analysing long documents — upload a PDF and ask detailed questions. Best for document analysis.

ChatGPT with Browse

Can search the web in real-time. Good for general research with current information.

Research Workflows

Summarising a Long Document

  1. Upload the document (PDF, paste text, or share a link)
  2. Ask for a high-level summary first: "Summarise the key points of this document in 5 bullet points"
  3. Then drill into specific sections: "What does section 3 say about [topic]?"
  4. Extract structured data: "Create a table of all recommendations with their rationale"

Learning a New Topic

  1. Start broad: "Explain [topic] like I'm a smart 15-year-old"
  2. Get the landscape: "What are the 5 most important things to understand about [topic]?"
  3. Go deeper: "Now explain [specific aspect] in more detail"
  4. Test understanding: "I think [your understanding]. Am I right, or what am I missing?"
  5. Get resources: "What are the best books/articles/courses to learn more about this?"

Competitive Analysis

"I run a [type of business] in [location]. My main competitors are [list them]. Analyse their strengths and weaknesses compared to us. Present as a comparison table with columns for: Company, Strengths, Weaknesses, Pricing, and Unique Selling Point. Use information you know about these companies — and clearly flag anything you're not sure about."

Common Mistakes

Don't

  • Trust AI statistics without checking
  • Use AI citations without verifying they exist
  • Ask one model and treat it as definitive
  • Rely on AI for medical, legal, or financial advice

Do

  • Cross-reference key claims with real sources
  • Ask AI to flag things it's uncertain about
  • Use AI to find leads, then verify yourself
  • Prefer tools that cite sources (Perplexity, Gemini)
All Guides