Writing & Editing Prompts

Source-to-Claim Attribution Editor

Align claims with their supporting sources, preserve uncertainty, and flag overreach, missing attribution, or evidence that does not support the wording.

FreeClaudeChatGPT
Best for

Writers, editors, analysts, and communications teams publishing evidence-based material that needs defensible claim wording and attribution.

Suitable LLM groups
FrontierReasoning
Download Prompt.md
Prompt
You are an evidence editor who specializes in claim precision, attribution, and source-faithful writing.

Inputs:
1. Draft text with citations or source markers: {{draft}}
2. Source excerpts, notes, or summaries: {{sources}}
3. Audience, publication type, and citation style: {{publication_context}}
4. Required claims, quotations, and disclosure rules: {{requirements}}
5. Fact-checking, access, length, and editorial constraints: {{constraints}}

Do the following:
1. Extract every factual, numerical, causal, comparative, predictive, and attributed claim, including claims implied by headings, captions, and transitions.
2. Match each claim to supplied source evidence and classify support as direct, partial, contextual, contradictory, missing, or unverifiable from the provided material.
3. Check quotation accuracy, attribution scope, date, population, geography, measurement, denominator, correlation-versus-causation language, and whether a source is primary or secondhand.
4. Revise overbroad claims to the narrowest wording supported by evidence, preserve material uncertainty and disagreement, and insert citation or verification markers without fabricating sources.
5. Produce the edited draft, claim-evidence matrix, unsupported-claim list, quotation check, and fact-check questions. Clearly flag sources that were supplied only as summaries and therefore could not be inspected directly.

How to use

  1. Include source excerpts with dates and context.
  2. Keep citation markers attached to the relevant claims.
  3. State the publication’s evidence standard.
  4. Have a specialist verify high-stakes claims.

Example input

Draft: 1,100-word article claiming remote work “increases productivity by 20%” and “reduces employee turnover.” Sources: one company pilot with 84 employees reporting 18% more completed tickets over eight weeks, a cross-industry survey showing lower stated intent to leave, and a review noting mixed results by job type. Audience: operations leaders. Style: linked sources in prose. Constraint: retain the article’s practical focus but remove causal certainty unsupported by the evidence.

Example output

The editor changes the productivity claim to a specific pilot result and adds company size, task measure, and duration. It replaces “reduces turnover” with “was associated with lower reported intention to leave” and distinguishes intention from actual turnover. The mixed review becomes the framing source for job-type variation. The claim matrix marks both causal conclusions as unsupported and lists missing evidence on selection effects and longer-term outcomes.

Customization tips

  • Check headings and captions for implied claims.
  • Preserve sample size, population, and time period.
  • Distinguish measured outcomes from self-reports.
  • Use verification markers instead of invented citations.

Tags

#claim-verification#source-attribution#fact-checking#evidence-editing#citation-review

FAQ

What is this prompt for?
It checks whether each claim and quotation is supported by the sources provided and revises overstatement.
How should I customize it?
Provide the draft, accessible source material, citation style, audience, required claims, quotations, and fact-checking constraints.
Are there any limitations?
It cannot verify inaccessible, summarized, fraudulent, or incomplete sources and does not replace specialist review for high-stakes publications.
How is it different from a basic prompt?
It builds a claim-evidence matrix and checks scope, attribution, population, dates, denominators, quotation accuracy, and causal language claim by claim.
Free

Translation Brief and Bilingual QA Editor

Prepare a translation brief and review a source-target pair for meaning, terminology, tone, locale fit, formatting, and functional accuracy.

ClaudeChatGPT
#translation-qa#localization#bilingual-editing
Free

Argument Strength Auditor

Reviews persuasive or opinion writing for logical gaps, weak evidence, and unsupported claims, then suggests specific edits to strengthen the argument.

ClaudeChatGPT
#persuasive-writing#critical-thinking#editing
Free

Inclusive Language and Bias Editor

Review writing for unnecessary stereotypes, exclusion, coded assumptions, accessibility barriers, and imprecise group references while preserving relevant meaning.

ClaudeChatGPT
#inclusive-language#bias-editing#editorial-review

Related Skills

Agent Systems & LLM WorkflowsFree

Agent Output Evidence and Citation Verifier

Verify agent outputs against supplied evidence using claim extraction, source matching, contradiction checks, confidence, unsupported-claim handling, and review gates.

ClaudeChatGPT
#agent verification#citation checking#LLM evidence
Testing & Quality ChecksFree

Evidence Claim and Citation Integrity Reviewer

Review claims, citations, quotations, statistics, and evidence alignment for accuracy, support, caution, and traceability.

ClaudeChatGPT
#citation review#evidence integrity#claim verification

Related Articles

Article · AI Agents

How do I build a support bot that does not hallucinate?

A support bot should answer only from approved sources, admit uncertainty, escalate edge cases, and avoid inventing policies, refunds, or technical fixes.

Jul 9, 20268 min read
Read How do I build a support bot that does not hallucinate?