Squeeze My Prompt
All tasks

I want to do something else.

Every question is optional. The ones marked Recommended matter most.

For example: Compare three phone plans for my family.

For example: Pick one by Friday and switch.

For example: 4 people, about 40GB a month, rural area, current plan $160.

For example: A table, then a 2-sentence recommendation.

For example: Only plans with no contract.

For example: Me and my partner.

For example: your pasted sample table.

More detail, if you have it

How should the AI start?
Saved on this device only

How to brief an AI for any other task

Use this form when none of the others fits: comparing options, planning a trip, sorting out a schedule, turning notes into a checklist. Its four recommended questions cover the core that the three largest AI makers agree a strong prompt contains: what to do, why, what the AI cannot guess, and what a good result looks like.

  • Task. Say what you want the AI to do, starting with a verb: compare, plan, summarize, draft. Google’s prompting guide advises clear and specific instructions, and its advice for its newest models is to state your goal clearly and concisely.
  • Purpose. Say what the result is for and what you will do with it. Anthropic’s prompting guide says that when you give the reason behind a request, the AI can apply it to choices you did not spell out.
  • Background. Give the facts the AI cannot guess, and paste anything relevant. Anthropic’s test is a good one: if a colleague with little context would be confused by your prompt, the AI will be too. Google’s prompting guide likewise advises supplying context rather than leaving the model to assume it.
  • What great looks like. Describe the ideal result: its shape, length and level of detail. Anthropic and OpenAI’s prompt guidance both advise describing the output you want and how you will judge it.

If you have an example of what you want, the optional “Example” question takes it. Examples steer an AI strongly, and it will follow yours closely, so use one when format or tone matters.

A worked example

A family is choosing a phone plan. Their answers: the task, the deadline it serves, their usage and budget, the result they want, and a sample of the comparison they have in mind. The form composes:

I need you to: Compare three phone plans for my family.

<context>
What this is for: Pick one by Friday and switch.
Background: 4 people, about 40GB a month, rural area, current plan $160.
</context>

<examples>
<example>
Plan A: 20GB, $40, no contract. Plan B: 50GB, $55, no contract.
</example>
This is an illustration of what I want, not a template to copy exactly.
</examples>

<output_format>
The ideal result: A table, then a 2-sentence recommendation.
</output_format>

If the information here is not enough to answer well, say so and tell me what you would need. If something missing would materially change the result, ask me up to 3 short questions first. Otherwise proceed and state your assumptions.

What the form adds for you

When you give an example, it tells the AI the example is an illustration and not a template, because Google’s prompting guide warns that models can copy examples too closely.