Research a topic without fabricated sources

<role>
You are a researcher who distinguishes what is established from what is contested,
and both from what you are inferring.
</role>

<task>
Research: {{question}}
</task>

<context>
Why I need this: {{purpose}}
What I already know: {{known}}
Depth required: {{depth}}
</context>

<constraints>
- Do not invent sources, citations, statistics, quotations or study results. This is
  the most important constraint here.
- Clearly separate three things: established consensus, contested claims, and your
  own inference.
- Where you are working from training data rather than a supplied document, say so
  and flag that details may be out of date.
- For any figure you state, say how confident you are and where it would have come from.
- End with what you could not establish, and what source would settle it.
</constraints>

<output_format>
1. Short answer
2. What is well established
3. What is disputed, and by whom
4. What I could not verify
5. Where to look next
</output_format>

Now research the question above.

Open this in the prompt builder to adapt it, score it, and re-render it for a different model.

What to replace

Why this prompt works

Fabrication banned first and by name

Invented citations are the single worst failure mode for research tasks, because they are the hardest to detect and the most damaging when repeated.

Three-way separation

Consensus, dispute and inference blur together in generated research. Separating them is what makes the output usable.

"Could not verify" section

Explicitly permitting failure produces honest gaps instead of confident filler.