Skip to content

Chapter 7

Brief it: ask clearly and give it what it needs


The chapter in a minute

  • A good request has four parts: a goal, the context it needs, an example when one helps, and the form of the answer.
  • Say who the answer is for and how firm or friendly to be, and paste in your own notes so it has nothing to guess.
  • When an answer misses, read it against your brief, find the part that let it down, fix that part and ask again, keeping what already worked.
  • Your prompt notebook keeps each prompt that worked, with the task, what worked, what to change and the date, so the next similar job starts from a tested brief.
  • Checklist question 2 is 'Have I given my goal, the context it needs, an example if it helps, and the form I want the answer in?'

Try it now

Same task, three briefs

0 of 6 steps done

Step 1 of 6

Choose one real writing task for this week, such as a thank-you note, a message to a club or a short notice. Keep private details out: first names only, and no addresses, phone numbers or account numbers.

Progress is saved in this browser only.

Prompts ready to go

The bare request

Write a letter to my landlord.

The full brief

Write a letter to my landlord asking her to send a plumber to fix a leak in my kitchen ceiling. I've emailed her three times with no reply, so I want it polite but firm.

Here are my notes. The leak is above the cooker and started on 9 October. It drips every night and fills a saucepan by morning. I emailed on 11, 15 and 19 October. I'm in any weekday after 4pm. Use only these facts.

Here is a line that shows the tone I want: 'I know you're busy, and I'd be grateful for a quick reply.' Keep it under 150 words, and end by asking her to reply by Friday.

Try this, from the book

Think of an answer from a chat assistant that missed for you lately. Write one line saying what was wrong and one line saying what you want instead. Then look at which of the four parts those two lines add.

Tools and settings, as of a date

The book points to the ‘Prompt library’ section of the companion page. Here it is in full, with the date each detail was checked. Open it on the companion page

Chapter 7 points here. The three sub-sections after it, "Prompt library: writing", "Prompt library: learning" and "Prompt library: planning", are named by chapters 11, 12 and 13.

Makers' own guides to writing prompts

The chapter teaches four parts of a good request: goal, context, example and the form you want the answer in. Each maker uses its own names for similar parts.

Maker's guideIts own names for the partsPage and date
Microsoft, "Get started writing prompts in Microsoft Copilot"Goal ("What do you want Copilot to do?"), Context ("Why do you need it, and who is involved?"), Source ("What information or samples should Copilot use?"), Expectations ("How should Copilot respond?"). A clear goal is all you need, and the other parts are optional.Microsoft Support (opens in a new tab). Checked 2 November 2026. Page shows "Last updated: September 2026".
Google, "How to write effective prompts"Four areas: Persona, Task, Context and Format. "You don't need to use all four in every prompt".Google Workspace (opens in a new tab). Checked 2 November 2026. Undated page.
Anthropic, "Prompting best practices" (developer guide)Be specific about the output format and constraints you want; add context; use examples, which are "one of the most reliable ways to steer" format, tone and structure.Claude Docs (opens in a new tab). Checked 2 November 2026. Undated page.
OpenAI, "Prompt engineering" (developer guide)Include additional context information; give a handful of input and output examples; precise instructions.OpenAI developer docs (opens in a new tab). Checked 2 November 2026. Undated page.

Two pages were not reachable when the source notes were made and are left out: one maker's prompting guide PDF and another's consumer help article on prompting.

One maker's guide, written for developers, advises softening forceful wording such as all-capital warnings, because newer models may over-react to it (Anthropic's guide, section "Tool usage"). That is developer advice, so treat it as background only.

Do magic words work?

The chapter says there are no reliable magic words. The studies behind that, with their limits:

Newer studies from 2026 point the same way:

  • Cai and colleagues, "Does Tone Change the Answer?" (first posted 14 December 2025, revised 27 March 2026, preprint): three models, very polite, neutral and very rude versions of the same questions. Rude wording lowered accuracy on some humanities questions for two of the models, but across whole subject areas the effects mostly disappeared. arXiv 2512.12812 (opens in a new tab)
  • Mehta and colleagues, "No Universal Courtesy" (17 April 2026, preprint): 22,500 prompts and answers, five models, three languages (English, Hindi and Spanish). Polite wording helped on average, but how much depended on the language and the model. arXiv 2604.16275 (opens in a new tab)

The effects of tone and phrasing vary by model. Checked: 2 November 2026.

A starter set from the book

These are the briefs the chapter uses, ready to copy. Change the details to your own, and keep the shape: goal, context, example, form.

The bare request and the full brief (chapter 7)

From the book

A brief has four parts: goal, context, example and form. Give the goal every time, and add the other three as the job gets more particular.
Your First Rung, chapter 7

Sources

  1. Prompting best practices (opens in a new tab) · Anthropic, 2026
  2. Language Models are Few-Shot Learners (opens in a new tab) · Brown, Tom B. and Mann, Benjamin and Ryder, Nick and Subbiah, Melanie and Kaplan, Jared and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, Tom and Child, Rewon and Ramesh, Aditya and Ziegler, Daniel M. and Wu, Jeffrey and Winter, Clemens and Hesse, Christopher and Chen, Mark and Sigler, Eric and Litwin, Mateusz and Gray, Scott and Chess, Benjamin and Clark, Jack and Berner, Christopher and McCandlish, Sam and Radford, Alec and Sutskever, Ilya and Amodei, Dario, 2020
  3. Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (opens in a new tab) · Dobariya, Om and Kumar, Akhil, 2025
  4. How to write effective prompts (opens in a new tab) · Google, 2026
  5. Large Language Models Understand and Can be Enhanced by Emotional Stimuli (opens in a new tab) · Li, Cheng and Wang, Jindong and Zhang, Yixuan and Zhu, Kaijie and Hou, Wenxin and Lian, Jianxun and Luo, Fang and Yang, Qiang and Xie, Xing, 2023
  6. Prompting Science Report 1: Prompt Engineering is Complicated and Contingent (opens in a new tab) · Meincke, Lennart and Mollick, Ethan and Mollick, Lilach and Shapiro, Dan, 2025
  7. Prompting Science Report 3: I'll pay you or I'll kill you - but will you care? (opens in a new tab) · Meincke, Lennart and Mollick, Ethan and Mollick, Lilach and Shapiro, Dan, 2025
  8. Get started writing prompts in Microsoft Copilot (opens in a new tab) · Microsoft, 2026
  9. Prompt engineering (opens in a new tab) · OpenAI, 2026
  10. Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting (opens in a new tab) · Sclar, Melanie and Choi, Yejin and Tsvetkov, Yulia and Suhr, Alane, 2024
  11. Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance (opens in a new tab) · Yin, Ziqi and Wang, Hao and Horio, Kaito and Kawahara, Daisuke and Sekine, Satoshi, 2024