Deep research prompt builder for a brief that supports a real decision
Turn a broad topic into a precise deep research brief with a decision, boundaries, source hierarchy and approval step.
- Works in
- ChatGPT, Claude, Gemini
- You need
- The decision you need to make · Known limits · Your geography and deadline · Required comparison criteria
- Written for
- deep research prompt template
A long research prompt is not automatically a good brief. The useful parts are the decision, scope, evidence standard and stopping conditions. This builder makes those visible before the expensive work begins.
See the full ChatGPT deep research guide for the review workflow that follows the brief.
Build a deep research brief. Do not conduct the research yet.
Start by restating the decision in one sentence. Then identify ambiguity that
could change the result. Ask no more than seven questions, ordered by how much
each answer would change the research plan.
After I answer, produce a brief with these sections:
Decision to support
Intended reader
In scope
Out of scope
Geography and date boundary
Required comparison criteria
Decision thresholds
Research questions
Source hierarchy
Sources to exclude or use only for discovery
Missing private inputs
Required output
Claim ledger fields
Conditions that would make the conclusion uncertain
The source hierarchy must name source types appropriate to the question. Put
primary sources first. Require dates and require repeated statistics to be
traced to their origin.
End with a research plan listing each sub question, the source type likely to
answer it and the evidence needed. Stop and ask for approval. Do not begin the
full research until I approve the plan.
Topic: [BROAD TOPIC]
Decision: [DECISION]
Reader: [WHO WILL USE THE REPORT]
Known boundaries: [GEOGRAPHY DATE BUDGET AND EXCLUSIONS]
Mandatory criteria: [CRITERIA]
Existing evidence: [WHAT WE ALREADY KNOW]What to change
Everything in square brackets is yours to replace. Nothing else needs editing.
[BROAD TOPIC]- The field the research will search, stated plainly.
[DECISION]- The choice the finished report must help somebody make.
[WHO WILL USE THE REPORT]- This controls depth, vocabulary and the form of the recommendation.
[GEOGRAPHY DATE BUDGET AND EXCLUSIONS]- These limits stop the research from silently choosing its own market and assumptions.
[CRITERIA]- List the features or measures every option must be judged against.
[WHAT WE ALREADY KNOW]- Existing data prevents repetition and reveals private inputs the public search cannot supply.
How to run it
- 01Write the decision first
If you only have a topic, decide who needs to choose what before using the prompt.
- 02Answer the clarification questions
Unknown boundaries become hidden assumptions later, so settle the important ones now.
- 03Inspect the source hierarchy
Make sure official records and original research outrank summaries where primary evidence exists.
- 04Approve the plan
Correct missing sub questions before spending time on the full research run.
Questions people ask
Why does the prompt stop before doing research?
A wrong scope contaminates the whole report. Reviewing the plan is the cheapest point to correct it.
What is a decision threshold?
It is a condition that changes the answer, such as a maximum price, minimum market size or mandatory feature.
Unsubscribe in one click. We never pass your address on.