Skip to content
PromptifyLab

Research a question without losing the sources

By Aamir1 min read

Use AI to discover and organise evidence, then separate what a source says from what you infer.

Before you start

A bounded research question, the time period that matters and a definition of a reliable source for your topic.

The workflow

1. Define the question before searching

Write the decision or explanation you need to support. Set a date range and scope. “What changed in this policy since last year?” is easier to verify than “tell me everything.”

2. Discover primary material

Use an AI search service to find promising sources. Prioritise the original research, official documentation or first-party records when they directly answer the question.

3. Open and inspect the evidence

Check the publication date, context and exact passage. A citation attached to a sentence does not prove that the source supports it. Mark inaccessible or ambiguous references as unverified.

4. Build an evidence table

Separate the claim, supporting passage, source URL, date and limitations. Draft from this table and keep disagreements visible. Do not make a consensus where the evidence is mixed.

A starting brief

Replace the bracketed fields. Paste it into ChatGPT, Claude, Gemini or another assistant – or press Fill in.

Starter prompt
Help me investigate: [QUESTION]. Scope and period: [SCOPE].

Return a table with claim, primary source URL, publication date, relevant passage to verify and limitations.

Separate source statements from your inferences.

Do not invent citations. If you cannot access a source, mark it unverified.

Finish with the most important unanswered question.

Open in

Make it concrete

Useful evidence row: “The feature supports X” → the official documentation page → the relevant section → date checked → any plan or platform restrictions.

Check the result

A reader should be able to trace the important claims back to accessible source material. Remove a claim if the evidence cannot support it.

Keep going

Sources

Product capabilities are based on these official pages. Workflow steps and sample briefs are editorial suggestions, not benchmark results. Checked 25 September 2026.