Fact-Checking and Source Audit

Check a draft claim by claim against the sources you paste, and get each one marked supported, partly supported, unsupported or needing a recheck.

Works in ChatGPT, Claude and Gemini9 of 20 in Content writing promptsUpdated 18 Oct 2026
  • writing
  • content
  • research
  • intermediate

The prompt

Prompt · 4 blanksBlanks in capitals
ROLE
Act as an experienced editor and research-led content strategist.

OBJECTIVE
Check DRAFT claim by claim against SOURCES and return supported, partly supported, unsupported, or needs-current-check statuses. Produce a practical fact-checking and source audit tailored to the user's context, not a generic overview.

INPUTS
- DRAFT
- SOURCES
- CONTEXT AND CONSTRAINTS
- SOURCE MATERIAL OR DATA (paste what is available; state ‘not supplied’ if none)

METHOD
1. Identify the intended user, goal, scope, and constraints before drafting.
2. Create a claim/evidence table with source, date or recency where available, confidence, and unresolved gaps.
3. Prioritize primary or authoritative sources; never fabricate a citation or imply you opened a source you could not inspect.
4. Make the reasoning traceable: connect recommendations to supplied evidence, examples, or explicitly labeled hypotheses.
5. If critical information is missing, ask up to three focused questions; otherwise proceed and label assumptions.

ACCURACY, PRIVACY, AND QUALITY RULES
- Never invent facts, figures, quotations, citations, credentials, customer results, policies, or URLs.
- For current or source-dependent claims, verify using reliable sources if accessible; cite only sources actually consulted. If verification is unavailable, list the checks needed.
- Protect confidential and personal information; use only details needed for the task. Follow applicable platform rules and user constraints.
- Avoid unsupported guarantees. Separate facts, estimates, interpretations, and recommendations.
- Prefer specific, usable output over filler, repetition, jargon, or vague advice.

OUTPUT FORMAT
- Executive summary and scope
- Findings with evidence or examples
- Prioritized issues and recommendations
- Assumptions, limitations, and validation needed
- Action plan with effort/impact where possible

FINAL SELF-CHECK
Before responding, check that the deliverable addresses ‘Fact-Checking and Source Audit’, uses the supplied context, labels unknowns, and can be reviewed or edited by a human. State any unresolved verification needs plainly.
Insert with the extension

Fill in the blanks

Type your details to see them in the prompt above, then copy it. What you type stays in this tab: it isn't saved or sent anywhere.

How to use it

  1. Fill in the blanks above, or paste the prompt into your AI chat and replace the words in capitals there. How blanks work
  2. Save it to your library to use it again. Your copy keeps the blanks, so next time you only fill them in.
  3. Use it where you chat. With the Save My Prompts extension, pick the prompt inside ChatGPT, Claude or Gemini and fill the blanks there.
  • Content Refresh Auditor

    Audit an old article or draft for outdated claims, intent mismatch and missing sections, and get a prioritised refresh list with the effort for each.

  • E-E-A-T Evidence Enhancer

    Map every claim in a draft to proof, author experience or a stated limit, flag what's unsupported, and see what to add to make the piece credible.

  • Research-Backed Article Planner

    Plan an article before you write it: a claim-and-evidence table with sources, confidence and gaps, plus a phased plan for drafting and review.

  • Citation and Attribution Assistant

    Find every place a draft needs a source or attribution, with the kind of reliable source to look for. It never invents references or quotations.

  • Email promptsCold Outreach Email

    Write a first-touch email to a type of recipient using your context, honest claims and an easy ask, plus the facts to check before you send it.

  • Social media promptsSocial Analytics Performance Auditor

    Paste your post metrics and get them read against your goals, formats and sample size, with real patterns kept apart from claims about causes.