Day 13: Research Anything Faster
The Concept
Research has two failure modes. The first is the rabbit hole: you start looking into something and end up reading adjacent topics for three hours without a clear answer to the original question. The second is the opposite: you do a quick search, accept the first credible-sounding result, and act on information that turns out to be incomplete or wrong.
AI does not fully solve either problem — it introduces its own risk around factual accuracy. But used correctly, it provides something the alternatives often lack: a rapid, structured orientation that tells you what you actually need to look for and why.
The role AI should play in research
Think of AI as the first conversation with a well-read colleague who knows the topic broadly. They give you the lay of the land — the main concepts, the key debates, the important distinctions, and the things to verify before relying on them. They point you toward where to look next. They do not replace the looking; they make the looking faster and more targeted.
This is significantly better than a search engine for topics where you need context before you know what to search. It is also faster than reading five background articles to achieve the same orientation. The difference is that AI can be asked to explain, summarise, and prioritise in a single response — rather than you assembling that picture across multiple sources.
Where verification becomes essential
Research is exactly where AI's limitations matter most. Any research that will inform a real decision — a professional recommendation, a medical question, a legal matter, a significant purchase — needs verification against an authoritative source. AI training data has a cutoff, models can present outdated information confidently, and hallucination risk is higher on specific factual claims than on broad explanations.
Perplexity AI is the most useful research-specific tool because it cites its sources. You can follow a citation to the original document, article, or publication — something you cannot do with an unsourced AI response. ChatGPT and Claude are better for orientation and explanation; Perplexity is better when you need traceable sources for claims that matter.
The practical research flow
Use AI to get oriented — understand the topic, identify the key questions, and find out what you need to verify. Then go to primary sources: official websites, academic papers, reputable publications, specialist organisations. The combination is faster and more reliable than either approach alone. AI gets you to the right questions quickly; primary sources give you the answers you can trust.
Prompt of the day
Copy this into your AI tool and replace any bracketed placeholders.
Prompt
I need to research [TOPIC]. I am starting with [DESCRIBE YOUR CURRENT LEVEL OF KNOWLEDGE — e.g. completely new to this / have a general sense but need more detail]. Please: 1) Give me a clear orientation to this topic — the key concepts, main distinctions, and what I actually need to understand. 2) Tell me the most important questions I should be trying to answer. 3) Identify the parts of this topic where I should verify information from an authoritative source rather than relying on your answer. 4) Recommend where to look for reliable, up-to-date information.
Your 15-minute task
Pick a topic you need to research this week — something for work, a decision you need to make, or something you want to understand better. Run the prompt. Then take one specific claim from the response and verify it using a search engine or authoritative source. Note whether it holds up.
Expected win
A clear orientation to your research topic, the key questions to investigate, and a verification habit that stops you acting on confident-sounding information that turns out to be wrong.
Power user tip
Use Perplexity AI for research specifically because it cites its sources. When AI gives you a useful response, ask: 'Which of these claims are most important for me to verify, and what search terms would I use?' This builds a targeted verification list rather than checking everything.