Most people use AI for research the same way: ask a question, read the confident answer, and treat it as settled. It's fast, it's articulate, and it's exactly the wrong way to use it — because the one thing a language model cannot reliably do is be your source of truth. It can sound certain about things that are wrong, and it can invent citations that don't exist.

The problem isn't that AI is useless for research. It's that its real value isn't being the source — it's helping you frame the question, find and read sources, evaluate them, spot your own bias, and reason toward an honest conclusion. This library is built around that: AI as a research partner that sharpens your thinking, while you keep your hands on the evidence.

Here's how to actually use it.

Make it reason with your evidence, not from its memory

The single most important habit in AI research is to stop asking it to recall facts and start giving it the material to reason about. Compare these:

What does the research say about intermittent fasting?

Here are three studies I found on intermittent fasting. Help me compare their methods and findings, and tell me where they conflict.

The first invites a confident summary you can't verify — and might be subtly or completely wrong. The second uses AI for what it's genuinely good at: reasoning over evidence you've gathered and can check. Every prompt here that touches sources is built to work from what you provide, and the system prompts are written never to fabricate facts or citations. When something needs verifying, they tell you to verify it.

Frame the question before you gather anything

This is the step that saves the most wasted effort, and the reason the library leads with Frame the Question.

Vague questions produce vague research. Before you read a single source, get sharp about what you're actually asking and what would count as an answer:

Help me turn this into a researchable question.

What evidence would actually answer my question?

Is my research question actually answerable?

If you can't say what would settle the question, you'll read forever and conclude nothing. A well-framed question is half the research done.

Fix big problems before small ones

When you're investigating something, work in this order:

Frame — turn curiosity into an answerable question.
Find — locate good sources, including ones that disagree.
Read & evaluate — extract accurately, judge credibility.
Analyze — test the reasoning and the numbers.
Synthesize — combine into an honest picture.
Conclude — claim only what the evidence supports.

The order matters because there's no point synthesizing sources you haven't evaluated, or concluding from evidence you haven't tested. Most people skip straight to an answer and then look for support — which is how research becomes the dressing-up of a foregone conclusion.

Evaluate before you believe

AI can summarize a source beautifully and tell you nothing about whether to trust it. The Evaluate Sources prompts are where the real rigor lives:

Help me assess whether this source is credible.

How strong is this evidence, really?

Is this really causation, or just correlation?

A claim's clarity is not its truth, and a well-written study can still be junk. Run the evidence through these before it earns a place in your thinking — and remember the AI can help you reason about credibility, but it can't confirm a fact for you.

Guard against the most dangerous bias — your own

The hardest bias to catch in research is the one you bring: the conclusion you wanted before you started. The Stay Honest prompts exist to turn the scrutiny inward:

Where might my own bias be affecting this research?

Am I cherry-picking evidence here?

Make the strongest case that I'm wrong.

That last one is the most valuable prompt in the whole library. If you can have AI build the best case against your conclusion and it still holds, you've got something. If it crumbles, better to find out now than in public.

A 30-second router

Not sure where to start? Match your situation:

Vague curiosity, no clear question → Frame the Question.
Need to find good information → Find Sources.
Have sources, need to digest them → Read & Extract.
Not sure what to trust → Evaluate Sources.
Testing an argument or claim → Analyze & Reason.
Lots of sources, no clear picture → Synthesize.
Planning a study or survey → Method & Design.
Working with numbers → Data & Numbers.
Writing it up → Write & Communicate.
Want to check you're being honest → Stay Honest.

A quick example

Say you want to know whether a popular supplement works, and you ask AI directly. It gives you a confident, balanced-sounding answer. You could stop there — and you'd have no idea if any of it is true.

Don't. Reframe: "what evidence would actually answer this?" Now you know to look for randomized trials, not testimonials. You find three studies and feed them in: "compare these methods and findings." One has forty participants and no control group — so you run "how strong is this evidence?" and discover it proves little. You synthesize the rest honestly, then run "make the strongest case I'm wrong" before concluding. You end up with a calibrated, defensible answer — and you can show your working.

The rigor didn't come from trusting the AI's first answer. It came from using it to frame, gather, evaluate, and stay honest — while you stayed in charge of the truth.

Don't want to choose? Follow the journey

If the full list is daunting, run the six-step journey instead — it walks you through an honest inquiry in the order that works:

Turn curiosity into a question — get something answerable.
Where should I look? — find the right sources.
Summarize this source — digest what you find.
Is this source credible? — judge before you trust.
Synthesize my sources — build the honest picture.
Draw a defensible conclusion — claim only what's supported.

Run those six in order and you've used the whole library's logic without browsing a single category.

The bottom line

Prompts are starting points; your evidence and your honesty do the heavy lifting. The best results come from three habits, every time:

Give it sources to reason about — don't use it as the source.
Frame what would answer the question before you gather anything.
Turn the scrutiny on your own conclusion, hardest of all.

If you only ever use ten of these, use these:

Turn curiosity into a researchable question.
Where should I look for good sources?
Summarize this source accurately.
Is this source credible?
How strong is this evidence?
Reason through this carefully.
Synthesize my sources.
Design my study.
Where might my own bias be affecting this?
How confident should I actually be?

Work the order, keep your hands on the evidence, and these prompts will do more than answer questions — they'll help you reach conclusions you can actually stand behind.

One caution worth repeating: an AI can help you think about sources, but it cannot verify facts or guarantee a citation is real. Always check primary sources yourself.