When you paste research sources into AI, the risk is that the model smooths over gaps. A verification prompt should force a claim-by-claim table and label what each source can and cannot prove.
Use it for
- Checking whether a source actually supports a claim.
- Separating facts, interpretations, and assumptions.
- Building a research log from pasted source notes.
- Finding claims that still need primary-source verification.
Copyable starter prompt
You are helping me verify sources for a student research project.
Use only the source notes I paste below. Extract each claim into a table with:
1. the exact claim,
2. the source that supports it,
3. the strongest quoted or paraphrased evidence,
4. what the source does not prove,
5. the confidence level,
6. the next source I should check.
Flag any claim that is unsupported, ambiguous, stale, or based on a same-name result. Do not smooth uncertainty into a stronger conclusion.What good verification looks like
The useful output is a claims table, not a polished paragraph. Each row should make it obvious whether the source proves the claim directly, only implies it, or merely gives you a lead. For school assignments, this is also the cleanest way to avoid accidental overclaiming.
After the table, ask the model for a short list of missing primary sources. If the topic is historical, legal, medical, financial, or tied to a current organization, treat secondary summaries as leads until you have checked the original source.