Claim Provenance Chain and Citation Traceability Audit
Audit consequential claims from current citation back to original source evidence, documenting where provenance breaks, mutates, or overstates authority.
Use in AI
Choose an AI tool to copy the current Prompt with a short usage note. Nothing is sent to that tool.
Objective: Trace each material claim through its full citation chain to the earliest available original source, dataset, record, or primary evidence. Identify where citations merely repeat intermediaries, where meaning changes across the chain, where authority is unsupported, and what release or correction decision is justified. Inputs to use: - Material claims to audit: [Material claims to audit] - Cited sources and intermediaries: [Cited sources and intermediaries] - Decision context: [Decision context] - Source access constraints: [Source access constraints] - Accountable release owner: [Accountable release owner] Instructions: 1. Treat each claim as a separate audit unit. Do not collapse similar claims unless the evidence chain is demonstrably identical. 2. For every claim, start with the provided citation or cited authority, then trace backward through cited papers, reports, datasets, webpages, legal records, interviews, statistics, or quoted statements until you reach the earliest accessible source that materially supports, weakens, contradicts, or fails to support the claim. 3. Use Perplexity search to locate originals and intermediaries where available. Prefer primary sources, original datasets, official records, archived pages, full papers, methodological appendices, and author-provided data over summaries, media reports, abstracts, AI-generated pages, and circular citations. 4. Do not assume that a cited source supports the claim because it is referenced. Verify the exact passage, table, figure, dataset field, legal clause, quote, date range, population, geography, method, and scope relevant to the claim. 5. Distinguish observations from inference: - Observation: what the source explicitly says or contains. - Inference: what must be concluded from comparing sources, methods, dates, or wording. - Unknown: what cannot be verified from accessible evidence. 6. Identify transformation points in the citation chain, including paraphrase drift, scope expansion, causal overstatement, missing denominators, outdated data reuse, quote truncation, statistical re-labeling, citation laundering, broken links, inaccessible originals, and claims supported only by secondary repetition. 7. Preserve uncertainty. If the original source is paywalled, missing, archived only, unavailable, or ambiguous, state that limitation and evaluate the chain only from available evidence. 8. Do not perform a generic source credibility ranking. Focus on whether the cited chain actually carries the audited claim from origin to current use without unsupported transformation. 9. Do not synthesize competing positions unless needed to evaluate whether a claim’s provenance is distorted. The central deliverable is claim-to-origin traceability. 10. Do not state that a source, dataset, archive, document, or full text was inspected unless you can cite the accessible evidence used. Deliver the audit in this format: A. Audit scope and evidence limits - Claims audited: - Sources or citation chains checked: - Sources not accessible or not located: - Material assumptions: - Consequence of unresolved gaps for [Decision context]: B. Claim-to-origin provenance graph For each claim, provide a compact chain using this structure: - Claim ID: - Current claim wording: - Current cited authority: - Intermediary 1: - Intermediary 2, if any: - Earliest accessible original source: - Evidence type at origin: - Direct support status: Supported / Partially supported / Not supported / Contradicted / Unverifiable - Provenance confidence: High / Medium / Low, with reason C. Citation transformation ledger Create a table with columns: - Claim ID - Chain step - Source and link - What this source actually states - How the next source or current claim changes it - Transformation type - Materiality of change: Low / Medium / High - Evidence quoted or summarized D. Source authority findings For each original or near-original source, assess only authority factors relevant to the claim: - Source identity and role in producing the evidence - Whether it is primary, near-primary, or secondary - Method, dataset, sample, jurisdiction, date range, or evidentiary basis - Known limitations stated by the source - Whether the current claim exceeds those limitations - Citation needed for the authority finding E. Broken-chain register List every provenance failure or unresolved gap: - Claim ID - Break or defect type - Location in chain - Why it matters - What evidence would repair the chain - Whether repair is feasible before release - Responsible owner: [Accountable release owner] or another named owner if more appropriate F. Release and correction decision For each claim, recommend one decision: - Release as written - Release with citation upgrade - Release with narrowed wording - Hold pending source repair - Remove claim - Issue correction or clarification if already published For each decision, include: - Required wording change, if any - Required citation replacement, if any - Residual risk - Verification step the [Accountable release owner] should complete before release Completion checks: - Every material claim has a traced chain or an explicit unresolved-chain entry. - Every releaseable claim links to an original or defensible near-original source. - Every unsupported transformation is recorded in the ledger or broken-chain register. - No unavailable source is described as inspected. - The final decision is usable by [Accountable release owner] for publication, release, correction, or withholding.
Variables to Replace
Replace each listed value in the Prompt with information relevant to your task.
- Material claims to audit
- Cited sources and intermediaries
- Decision context
- Source access constraints
- Accountable release owner
How to Use This Prompt
Use this in Perplexity. Paste the exact claims, existing citations, links, PDFs, bibliographies, draft excerpts, and any access limits. Replace every bracketed placeholder, then run the prompt. Afterward, have the accountable release owner or editorial owner verify any proposed wording changes, citation replacements, and unresolved provenance gaps before publication or correction.
Example Use Case
A research director is preparing a public report that cites several secondary studies for claims about health outcomes. The prompt traces each claim through news coverage, literature reviews, underlying studies, and datasets, revealing that two claims rely on paraphrase drift and one lacks access to the original dataset. The release owner narrows one claim, replaces two citations with primary sources, and holds one claim pending source repair.
Was this useful?