Prompt library — 15 prompts
Reading fast, checking claims, summarising sources and spotting what is missing.
Use this to organize research you've already gathered, not to generate claims about competitors from the model's own assumptions.
Here are my notes and screenshots from reviewing [COMPETITOR/S] content: [PASTE NOTES]. Synthesize the patterns: recurring themes, formats they use most, and any gap I might be able to fill. Flag clearly that this is based only on what I've provided — I'll verify current positioning directly before acting on it.
Use this to make sense of a pile of reading, distinguishing consensus from a single hot take.
Here are several articles/reports I've read on [TOPIC]: [PASTE EXCERPTS OR SUMMARIES WITH SOURCES]. Synthesize the common threads and where sources disagree. Clearly separate what multiple sources agree on (higher confidence) from what a single source claims (lower confidence, needs more verification).
Use this after collecting qualitative feedback, before it turns into an unread pile of notes.
Here are notes or transcripts from [NUMBER] interviews/survey responses on [TOPIC]: [PASTE]. Extract the recurring themes, note roughly how many respondents touched each theme, and pull one representative verbatim quote per theme.
Use this for a long report, contract, or research paper you need the substance of quickly.
Summarize this document: [PASTE OR ATTACH]. Give me: a 3-sentence summary, the 5 most important points as bullets, and anything that seems like it needs a decision or action from me specifically. Flag if the document contradicts itself anywhere.
Use this when researching a topic with genuinely mixed opinions, so you see the disagreement instead of a false consensus.
I have [NUMBER] sources on [TOPIC]: [PASTE OR SUMMARIZE EACH]. Compare them: where do they agree, where do they conflict, and what's each source's likely bias or angle (e.g. vendor-published, academic, opinion piece). Don't resolve the conflicts for me — show me where they are.
Use this when feedback has piled up and you need the pattern, not each comment individually.
Here's a batch of customer feedback (reviews, support tickets, survey comments): [PASTE]. Group it into themes, rank themes by frequency, and separate feedback that's about the product/service itself from feedback that's about price, support, or something adjacent.
Use this for longer meetings where decisions get buried in discussion.
Here's a meeting transcript: [PASTE]. Extract only what was actually decided (not proposed, not discussed) — for each decision, who made the call and any condition attached to it. Separate this from a list of things that were discussed but left open.
Use this to structure a SWOT from information you already have, not as a source of new market facts.
Build a SWOT based only on what I give you: strengths and weaknesses I've listed [LIST], and opportunities/threats from this market context [PASTE CONTEXT]. Don't invent strengths or threats I haven't given evidence for — if a quadrant is thin, tell me it's thin rather than padding it.
Use this to read your own numbers honestly, including when they don't support the story you were hoping for.
Here's my content performance data before and after [CHANGE — e.g. new posting format, new hook style]: [PASTE DATA]. Analyze whether the change correlates with a real difference, and flag if the sample size or timeframe is too small to draw a confident conclusion.
Use this for a decision with real trade-offs, where you want the reasoning visible, not just a recommendation.
I'm deciding between [OPTIONS]. Build a decision matrix with the criteria that actually matter for this decision [LIST YOUR REAL CRITERIA], weighted by importance to me [RANK OR WEIGHT THEM]. Score each option and show the math, so I can see where I disagree with a score rather than just trusting a final answer.
Use this before conducting user or market research interviews.
I'm researching [TOPIC] by talking to [WHO YOU'LL INTERVIEW]. Write 8 open-ended questions that avoid leading the answer, ordered from broad context to specific detail. Flag any question that's actually two questions at once.
Use this as a first pass to find what needs verification — always confirm the actual facts against a primary source yourself before publishing.
Review this draft for factual claims that need verification before publishing: [PASTE DRAFT]. List each specific claim (a stat, a named tool feature, a quote) separately so I can check each one against a primary source myself. Don't confirm or deny accuracy — just flag what needs checking.
Use this to quickly read a spreadsheet export without manually scanning every row.
Here's a data table: [PASTE TABLE]. Summarize what it shows in plain language — the headline pattern, the biggest outlier, and anything that looks like it might be a data quality issue rather than a real trend.
Use this to go beyond a star-rating average and understand what's actually driving sentiment.
Here are reviews/comments about [PRODUCT/SERVICE]: [PASTE]. Identify sentiment patterns — not just positive/negative, but what specifically drives strong positive reactions versus what drives complaints. Pull the most representative quote for each pattern.
Use this before starting research, so the question you're answering is actually answerable.
My research question is currently: [VAGUE QUESTION, e.g. 'do people like AI tools']. Help me sharpen it into something specific and answerable — what population, what specific behavior or belief, and what would count as evidence one way or the other.