TL;DR - Don't say "summarize" - say who it's for and why. A purpose-built summary turns hours of reading into minutes.
Why it matters
Knowledge work is reading before doing. AI compresses that cycle - but a generic summary is nearly useless. Tell it the reader and the goal and you get exactly what you need to act.
Worked example
Summarize this 8-page report for a busy exec:
- 3 bullets on what changed
- 1 bullet on the biggest risk
- 1 bullet on what we should do nextFar more useful than "summarize this report."
Steal this - extract patterns
"List every decision and who made it."
"Pull the 5 numbers a finance team would care about."
"Give me the action items with owners and dates."
"What are the 3 strongest arguments for and against?"Common mistakes (and the fix)
- Generic "summarize." Fix: state the reader + purpose.
- Trusting research as fact. Fix: treat AI research as leads; verify what you'll rely on.
- Forgetting the source's limits. Fix: a non-browsing chat doesn't know today's news.
Good to know
For sourced research use a tool that cites: Perplexity (live, footnoted answers) or NotebookLM (grounds answers in documents you upload - great for "ask my PDFs"). For summarizing long docs, Claude and Gemini handle very large inputs in one go.