Skip to content
Get in touch
Menu
Field Guide · level 2 · Everyday skills

Why does your prompt only make sense to you?

Add what only you know: the reader, and the use.

5 min · try it 5 min · as of Oct 2026
Bring this to your team · Team workshopsA workshop on your team's real requests: name each usual reader once, and share it. Get in touch
01

Fine, for nobody

“Summarise this report.” A fine summary came back, for nobody in particular.

02

Only you hear the tune

A 1990 study: tappers expected half their songs named, on average. Listeners got 1 in 40.

03

Your request is the tapping

Our comparison: it gets the taps, not the tune. So it picks a typical reader.

04

Three moves

Before you send:

  1. Who reads it
  2. How they'll use it
  3. Say it once
05

1. Who reads it

And what they know: “For my boss. She knows the project, not the figures.”

06

2. How they'll use it

Where and when they'll read it. “On her phone, two minutes before a meeting.”

My pick for writing

ChatGPT logoChatGPTAlso good: Claude logoClaude

Quick drafts in a chat, and ChatGPT for Word to work right inside your document.

All my picks
07

3. Say it once

Save your usual reader in standing instructions. Add the lines for anyone else.

08

Put the tune into words

Two lines in every request: who reads it, and how they'll use it.

Try it · 5 minutes

Catch its guess

Use made-up or public material.

  1. Take a request you send often.1 min
  2. Add: “First, guess who reads this, and how.”2 min
  3. Read its guess. Reply with the two lines.2 min

The two lines

For: [who reads it, and what they already know]
Use: [where and when they'll read it, and what they'll do next]

Good looks likeThe next answer fits your reader.

In your assistant · as of Oct 2026 · Claude, ChatGPT, Gemini

Say your usual reader once

Standing instructions go into every chat. Menus move often; at work, an admin may decide.

  • CheckYour role, in brief
  • CheckWho usually reads it
  • CheckHow they usually read

Use roles, not names: “my director”, not her name. Keep clients and anything private out of standing instructions.

Claude

Instructions for Claude, and project instructions

Every chat
SettingsthenInstructions for Claude

Click your initials first. Applies to all your chats.

One job
ProjectsthenSet project instructions

Only chats in that project use them. Save instructions when done.

Each time

Someone new? Add the two lines to the request.

For: who reads it. Use: how they'll use it.

ChatGPT

Custom Instructions, and project instructions

Every chat
SettingsthenPersonalizationthenCustom Instructions

Turn on Enable customization first. On every plan.

One job
•••thenProject settings

Project instructions apply only in that project and override your custom instructions.

Each time

Someone new? Add the two lines to the request.

For: who reads it. Use: how they'll use it.

Gemini

Instructions for Gemini, and skills

Every chat
Settings & helpthenPersonal IntelligencethenInstructions for Gemini

Then Add, type it and Submit. Gemini applies it to every chat.

One job
CustomizethenSkillsthenCreate

Saves instructions for one kind of task. Personal accounts, 18 or over, for now.

Each time

Someone new? Add the two lines to the request.

For: who reads it. Use: how they'll use it.

Can't find Customize? Click More in the sidebar first.

Quick check · three questions

Did it stick?

  1. 1Your request reads perfectly to you. Why can the answer still miss?

    The answer: B What's obvious to you never reaches it. Two plain lines fix that, not clever wording.
  2. 2You ask: “Write a note about the office move.” Which line adds the most?

    The answer: C It names the reader and the use. Adjectives and job titles leave the reader to a guess.
  3. 3You write for the same director most weeks. What saves retyping?

    The answer: A Standing instructions go into every chat. Add the two lines when the reader is someone else.
0 of 3 answered
The studySide note
Elizabeth Newton, Stanford, 1990: 120 songs tapped, 3 named. The tappers' guesses averaged 50 percent, from 10 to 95. It tested people, not AI; the comparison with AI is ours.
The vendors say it tooSide note
Anthropic's prompting article lists the audience, and how the output will be used, as context worth giving. OpenAI's help page says models often do best when you ask as you would ask a person.
The rest of a briefSide note
The reader and the use are two lines of a good brief. For the goal, the source and what good looks like, see How do you brief AI so the first answer is close?
Sources, checked 18 Oct 2026Source

Product details as of Oct 2026, each checked on the vendor's own page on 18 Oct 2026. Recheck by 18 Nov 2026. The examples in the pictures are made up for this lesson.

Free kitThe 2026 AI StarterBring this to your teamTeam workshops · Get in touchBack to level 2, Everyday skills