Squeeze My Prompt
All tasks

I want to research a topic.

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

For example: How have US office vacancy rates changed since 2020, and what's driving it?

For example: Deciding whether to sign a 5-year office lease in Denver next spring.

For example: 2020 to now; US only, with Denver detail; offices only.

For example: Prefer government data and broker reports; avoid press releases.

For example: One page: 5 findings, a table of vacancy by year, sources at the end.

For example: Full report.

  • Quick answer: a short answer in the chat
  • Full report: a long report from a deep research tool

For example: I've read vacancy is near 20% nationally. Check that.

For example: Me and my business partner; not property experts.

For example: Both 'office recovery' and 'structural decline'.

For example: My broker's market report.

For example: Not residential property.

More detail, if you have it

How should the AI start?
Saved on this device only

How to brief an AI to research a topic

Use this form when you need facts gathered and weighed: a market, a policy, a product category, a question you want answered with sources. It works for a quick answer in a chat assistant and for a full report from a deep research tool. Deep research tools build a plan from what you give them before they search, so the details you enter decide what gets looked at.

  • Question. Say exactly what you want to find out. “Clearly define your research objective” is the first of OpenAI’s three tips for deep research, and Perplexity’s help center puts it simply: a focused question leads to a focused report.
  • Purpose. What you will do with the answer turns a topic into a question the AI can answer. Anthropic’s prompting guide advises defining what a successful answer looks like.
  • Scope. Period, places and limits. OpenAI’s guide to deep research in its API tells its prompt rewriter to include every detail the user gives and to mark missing ones as open-ended rather than invent them. Stating the scope yourself leaves less open. Perplexity’s prompt guide warns that made-up answers are most likely when content is paywalled or very recent.
  • Sources. Name the sources you trust or want avoided. OpenAI’s guide says to specify priority sources, and Gemini Deep Research lets you limit a search to sources you choose.
  • Output. Say what the answer should look like. OpenAI’s guide asks for the expected output format, and Perplexity advises capping the length of lists.

A worked example

Someone deciding on an office lease asks how vacancy rates have moved. The form composes:

I want to find out: How have US office vacancy rates changed since 2020, and what's driving it?

<context>
What I'll do with the answer: Deciding whether to sign a 5-year office lease in Denver next spring.
</context>

<constraints>
Scope: 2020 to now; US only, with Denver detail; offices only.
Sources to prefer or avoid: Prefer government data and broker reports; avoid press releases.
Cite a source for each factual claim, with its date. If you find only a near match (a different year, a related organization), say so.
</constraints>

<output_format>
Layout: One page: 5 findings, a table of vacancy by year, sources at the end.
I want a full research report.
</output_format>

Today's date is November 10, 2026. 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

It asks for a dated source for each claim and a warning when only a near match turns up, because Perplexity’s prompt guide says made-up answers are also most likely when a search returns a near miss, such as a different year.