Prompt libraryResearch and decisions
Compare options, find what you do not know and prepare to verify before you decide.
Research prompts should ask for structure, uncertainty and what to verify, not for conclusions. An assistant that gives a confident verdict on a contested question is hiding the part you most need to see.
Treat the output as a reading list and a set of questions. Open the primary sources before relying on a claim, and write the date you checked.
Compare [option A] and [option B] for someone who [situation]. Use a table with criteria, what each does better, what each does worse, and one thing that would change the recommendation. Say where the evidence is thin.Why it works. Criteria and "what would change it" force a conditional answer.
Then check. Verify prices and specifications on official pages and note the date.
Describe the main positions on [question], the strongest argument for each, the evidence each side relies on, and where they disagree about facts versus values. Do not pick a side.Why it works. Separating facts from values clarifies the real disagreement.
Then check. Look for primary sources for the key evidence.
I'm trying to decide [decision]. Here is what I know: [facts]. List the five questions I should answer before deciding, and who or what would answer each.Why it works. Good questions are worth more than a quick answer.
Then check. Add questions from your own context.
For the claim "[claim]", what kinds of primary sources would settle it, which would be weak evidence, and what search terms would find them?Why it works. Directs you to evidence instead of more AI text.
Then check. Actually open the sources.
I plan to [plan]. Argue against it as strongly as you honestly can, then say which of your arguments is weakest.Why it works. Counters the tendency to agree with you.
Then check. Judge which objections hold.
Build a timeline of [topic]. Mark each entry as well-documented, disputed or uncertain, and say how recent the information is likely to be.Why it works. Labels prevent false precision.
Then check. Verify dates against reliable references.
Estimate [quantity] step by step. State each assumption and a plausible range, show the arithmetic, and say which assumption matters most.Why it works. Visible assumptions make an estimate criticizable.
Then check. Recompute the arithmetic with a calculator.