With a long document — a research report, a transcript, a year of social comments — order matters. Put the material at the top and your question at the very end.
Summarise the three biggest complaints in these comments: [5,000 pasted comments]
[5,000 pasted comments inside <comments> tags] — then, at the end: Read the comments above. What are the three biggest complaints, with how often each appears and two verbatim quotes for each?
The prompt
Copy-ready prompt
<documents> <document index="1"> <source>[FILE NAME OR SOURCE]</source> <document_content> [PASTE THE LONG MATERIAL HERE] </document_content> </document> </documents> Using only the documents above, answer: [YOUR QUESTION]. Quote the passages you rely on before giving your answer.
Why it works
Anthropic advises placing long documents “near the top of your prompt, above your query”, and reports that queries at the end can improve response quality by up to 30 percent in its tests, especially with multiple documents. Google’s Gemini guide gives the same rule: supply large context blocks first, questions last.
Use it for
Consumer research, focus-group transcripts, review mining, competitor reports. Asking the model to quote before it answers also makes it easy for you to check the answer against the source.
Sources
- Anthropic — Claude prompting best practices — platform.claude.com (accessed 28 Sep 2026)
- Google — Prompt design strategies for the Gemini API (updated 17 Sep 2026) — ai.google.dev (accessed 28 Sep 2026)
More tricks and templates in the Prompt Lab. Want this applied to your brand’s brief? Start a project.